Publish reproducible CNN baseline cnn-001

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+models/cnn_001_participant_disjoint.keras filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,60 @@
+---
+language:
+- en
+library_name: tensorflow
+pipeline_tag: image-classification
+tags:
+- asl
+- sign-language
+- cnn
+- reproducible-research
+---
+
+# ASL-HG CNN baseline (`cnn-001`)
+
+A TensorFlow/Keras CNN baseline for 36 static ASL classes (digits `0–9`, letters `A–Z`). This repository contains the trained checkpoint **and the full experiment record** needed to reproduce or audit the result.
+
+## Result
+
+| Metric | Value |
+|---|---:|
+| Test accuracy | 83.2544% |
+| Macro F1 | 80.0983% |
+| Macro precision / recall | 79.5670% / 83.3056% |
+| Best validation accuracy | 72.0964% |
+| Epochs run | 15 |
+| Recall `O` / `0` | 100.0000% / 0.0000% |
+
+This is a deliberately simple CNN-from-scratch baseline, not the final proposed model. In particular, digit `0` has zero recall in this run, so later improvements should report this class separately.
+
+## Evaluation protocol
+
+- Dataset: [`hnam25/asl-hand-gesture-images`](https://huggingface.co/datasets/hnam25/asl-hand-gesture-images), revision `8f36ac00ece6dfce94410a980a839d93a912d366`.
+- Input: the publisher's `ASL_Processed_Images.zip`, SHA-256 `a8e7a38c4085fd9dc18aa4fa8646ad7d6917e9ff7374a3ffd5f9f3b37a5c045b`.
+- Audit source: `metadata/colab-audit-2026-08-10`; the raw archive fingerprint and audit statistics are preserved in `metadata/experiment_config.json`.
+- Exact duplicate policy: retain one deterministic canonical image per raw-image SHA-256. The run removes 559 duplicate crops from 36000 usable images.
+- Split: **participant-disjoint**. Train: P1, P10, P3, P4, P5, P6, P7, P8; validation: P2; test: P9. No participant or exact image hash appears in multiple partitions.
+- Seed: 42. Split files and the deduplication manifest are included under `metadata/`.
+
+This participant-disjoint protocol is stricter than the publisher's supplied image-level train/test archive split and should not be numerically compared with a random image-level split.
+
+## Model
+
+`Rescaling(1/255) → 3 × [Conv-BN-ReLU-Conv-BN-ReLU-MaxPool] (32/64/128 filters) → GAP → Dense(256) → Dropout(0.3) → Dense(36, softmax)`.
+
+Training used Adam (`1e-3`), image size 128, batch size 64, at most 30 epochs, checkpointing on validation accuracy, early stopping on validation loss, and learning-rate reduction on plateau. Best validation accuracy occurred at epoch 10.
+
+## Files
+
+- `models/cnn_001_participant_disjoint.keras`: best checkpoint.
+- `metrics/`: summary, per-class report, confusion matrix.
+- `figures/confusion_matrix.png`: visual confusion matrix.
+- `logs/training_history.csv`: full optimization history.
+- `metadata/experiment_config.json`: pinned data/config provenance.
+- `metadata/split_manifest.json`, `train.csv`, `validation.csv`, `test.csv`, `deduplication_manifest.csv`: exact protocol and partitions.
+- `reproducibility/08_cnn_baseline_reproducible.ipynb`: Colab notebook used for this run.
+- `reproducibility/requirements.txt`: project requirements snapshot.
+
+## Reproduce
+
+Open the included notebook in Google Colab, run it on a T4 GPU, and use the pinned dataset revision already embedded in the notebook. It downloads public data and audit metadata, verifies the processed archive and audit mapping, recreates the exact participant split, trains, and writes the same artifact layout.

logs/training_history.csv

@@ -0,0 +1,16 @@
+accuracy,loss,val_accuracy,val_loss,learning_rate
+0.4591221511363983,1.765419363975525,0.08603384345769882,6.974681854248047,0.0010000000474974513
+0.8571126461029053,0.44405597448349,0.31000861525535583,2.6612038612365723,0.0010000000474974513
+0.9495857357978821,0.17806586623191833,0.423286497592926,3.7735824584960938,0.0010000000474974513
+0.9778600335121155,0.08823823183774948,0.49784916639328003,1.708282709121704,0.0010000000474974513
+0.98769611120224,0.05259506031870842,0.42730140686035156,2.4472529888153076,0.0010000000474974513
+0.9845231771469116,0.05536973848938942,0.5976483821868896,1.0407171249389648,0.0010000000474974513
+0.9936894178390503,0.02670445665717125,0.5190708637237549,2.467351198196411,0.0010000000474974513
+0.9900934100151062,0.035457175225019455,0.22053341567516327,8.213214874267578,0.0010000000474974513
+0.9988013505935669,0.008368562906980515,0.7097792029380798,1.056762933731079,0.00020000000949949026
+0.9996827244758606,0.0033378207590430975,0.7209635972976685,1.0305737257003784,0.00020000000949949026
+0.9997884631156921,0.0028094735462218523,0.6724978685379028,1.2147690057754517,0.00020000000949949026
+0.9998589754104614,0.0024102323222905397,0.6570117473602295,1.3460917472839355,0.00020000000949949026
+1.0,0.001786125940270722,0.696587324142456,1.0568524599075317,4.0000002627493814e-05
+0.9998237490653992,0.0018538410076871514,0.7069113850593567,1.065900206565857,4.0000002627493814e-05
+0.9999294877052307,0.0016534593887627125,0.7051907181739807,1.0694741010665894,8.000000889296643e-06

metadata/environment.txt

@@ -0,0 +1,703 @@
+absl-py==1.4.0
+accelerate==1.14.0
+access==1.1.10.post3
+affine==2.4.0
+aiofiles==25.1.0
+aiohappyeyeballs==2.7.1
+aiohttp==3.14.1
+aiosignal==1.4.0
+aiosqlite==0.22.1
+alabaster==1.0.0
+albucore==0.0.24
+albumentations==2.0.8
+ale-py==0.12.0
+altair==5.5.0
+annotated-doc==0.0.4
+annotated-types==0.7.0
+antlr4-python3-runtime==4.9.3
+anyio==4.14.2
+anywidget==0.9.21
+apsw==3.53.3.1
+apswutils==0.1.2
+argon2-cffi==25.1.0
+argon2-cffi-bindings==25.1.0
+array_record==0.8.3
+arrow==1.4.0
+arviz==0.22.0
+astropy==7.2.2
+astropy-iers-data==0.2026.7.13.0.54.2
+astunparse==1.6.3
+atpublic==5.1
+attrs==26.1.0
+audioread==3.1.0
+Authlib==1.7.2
+autograd==1.9.1
+babel==2.18.0
+backcall==0.2.0
+beartype==0.22.9
+beautifulsoup4==4.13.5
+betterproto==2.0.0b6
+bigframes==2.42.0
+bigquery-magics==0.14.0
+bleach==6.4.0
+blinker==1.9.0
+blis==1.3.3
+blobfile==3.2.0
+blosc2==4.8.0
+bokeh==3.8.2
+Bottleneck==1.4.2
+bqplot==0.12.47
+branca==0.8.2
+brotli==1.2.0
+CacheControl==0.14.4
+cachetools==6.2.6
+catalogue==2.0.10
+certifi==2026.6.17
+cffi==2.1.0
+chardet==5.2.0
+charset-normalizer==3.4.9
+clarabel==0.11.1
+click==8.4.2
+cligj==0.7.2
+cloudpathlib==0.24.0
+cloudpickle==3.1.2
+cmake==3.31.10
+cmdstanpy==1.3.0
+colorcet==3.2.1
+colorlover==0.3.0
+community==1.0.0b1
+confection==1.3.3
+cons==0.4.7
+contourpy==1.3.3
+cramjam==2.11.0
+cryptography==49.0.0
+cucim-cu12 @ https://pypi.nvidia.com/cucim-cu12/cucim_cu12-26.2.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
+cuda-bindings==12.9.7
+cuda-core==0.3.2
+cuda-pathfinder==1.5.6
+cuda-python==12.9.7
+cuda-toolkit==12.8.1
+cudf-cu12==26.2.1
+cudf-polars-cu12==26.2.1
+cufflinks==0.17.3
+cuml-cu12==26.2.0
+cupy-cuda12x==14.0.1
+curl_cffi==0.15.0
+cuvs-cu12 @ https://pypi.nvidia.com/cuvs-cu12/cuvs_cu12-26.2.0-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl
+cvxopt==1.3.2
+cvxpy==1.6.7
+cycler==0.12.1
+cyipopt==1.5.0
+cymem==2.0.13
+Cython==3.0.12
+dask==2026.1.1
+dask-cuda==26.2.0
+dask-cudf-cu12==26.2.1
+dataproc-spark-connect==1.1.0
+datasets==4.0.0
+db-dtypes==1.7.1
+dbus-python==1.2.18
+debugpy==1.8.15
+decorator==4.4.2
+defusedxml==0.7.1
+deprecation==2.1.0
+diffusers==0.39.0
+dill==0.3.8
+distributed==2026.1.1
+distributed-ucxx-cu12==0.48.0
+distro==1.9.0
+dlib==19.24.6
+dm-tree==0.1.10
+docstring_parser==0.18.0
+docutils==0.21.2
+dopamine_rl==4.1.2
+duckdb==1.3.2
+earthengine-api==1.7.35
+easydict==1.13
+editdistance==0.8.1
+eerepr==0.1.2
+einops==0.8.2
+en_core_web_sm @ https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-3.8.0/en_core_web_sm-3.8.0-py3-none-any.whl#sha256=1932429db727d4bff3deed6b34cfc05df17794f4a52eeb26cf8928f7c1a0fb85
+entrypoints==0.4
+esda==2.9.0
+et_xmlfile==2.0.0
+etils==1.14.0
+etuples==0.3.10
+Farama-Notifications==0.0.6
+fastai==2.8.7
+fastapi==0.139.0
+fastcore==2.0.5
+fastdownload==0.0.7
+fastjsonschema==2.21.2
+fastlite==0.2.4
+fastprogress==1.1.6
+fasttransform==0.0.2
+filelock==3.29.7
+firebase-admin==6.9.0
+Flask==3.1.3
+flatbuffers==25.12.19
+flax==0.11.2
+folium==0.20.0
+fonttools==4.63.0
+fqdn==1.5.1
+frozendict==2.4.7
+frozenlist==1.8.0
+fsspec==2025.3.0
+future==1.0.0
+gast==0.7.0
+gcsfs==2025.3.0
+GDAL==3.8.4
+gdown==5.2.2
+geemap==0.38.3
+geocoder==1.38.1
+geographiclib==2.1
+geopandas==1.1.4
+geopy==2.4.1
+giddy==2.3.6
+gin-config==0.5.0
+gitdb==4.0.12
+GitPython==3.1.51
+glob2==0.7
+google==3.0.0
+google-adk==2.4.0
+google-ai-generativelanguage==0.6.15
+google-api-core==2.30.3
+google-api-python-client==2.198.0
+google-auth==2.49.0
+google-auth-httplib2==0.4.0
+google-auth-oauthlib==1.4.0
+google-cloud-bigquery==3.42.2
+google-cloud-bigquery-connection==1.22.0
+google-cloud-bigquery-storage==2.39.0
+google-cloud-core==2.6.0
+google-cloud-dataproc==5.28.0
+google-cloud-datastore==2.25.0
+google-cloud-firestore==2.27.0
+google-cloud-functions==1.24.0
+google-cloud-language==2.21.0
+google-cloud-monitoring==2.31.0
+google-cloud-resource-manager==1.18.0
+google-cloud-spanner==3.68.0
+google-cloud-storage==3.13.0
+google-cloud-translate==3.27.0
+google-colab @ file:///colabtools/dist/google_colab-1.0.0.tar.gz
+google-crc32c==1.8.0
+google-genai==2.11.0
+google-generativeai==0.8.6
+google-pasta==0.2.0
+google-resumable-media==2.10.0
+googleapis-common-protos==1.75.0
+googledrivedownloader==1.1.0
+gradio==6.20.0
+gradio_client==2.5.0
+grain==0.2.18
+graphviz==0.21
+greenlet==3.5.3
+groovy==0.1.2
+grpc-google-iam-v1==0.14.4
+grpc-interceptor==0.15.4
+grpcio==1.82.1
+grpcio-status==1.71.2
+grpclib==0.4.9
+gspread==6.2.1
+gspread-dataframe==4.0.0
+gym==0.25.2
+gym-notices==0.1.0
+gymnasium==1.3.0
+h11==0.16.0
+h2==4.3.0
+h5netcdf==1.8.1
+h5py==3.16.0
+hdbscan==0.8.44
+hf-gradio==0.4.1
+hf-xet==1.5.1
+highspy==1.15.1
+holidays==0.100
+holoviews==1.23.1
+hpack==4.2.0
+html5lib==1.1
+httpcore==1.0.9
+httpcore2==2.5.0
+httpimport==1.4.1
+httplib2==0.32.0
+httptools==0.8.0
+httpx==0.28.1
+httpx2==2.5.0
+huggingface_hub==1.23.0
+humanize==4.16.0
+hyperframe==6.1.0
+hyperopt==0.2.7
+ibis-framework==9.5.0
+idna==3.18
+ImageIO==2.37.3
+imageio-ffmpeg==0.6.0
+imagesize==2.0.0
+imbalanced-learn==0.14.2
+immutabledict==4.3.1
+importlib_metadata==9.0.0
+importlib_resources==7.1.0
+imutils==0.5.4
+inequality==1.1.2
+inflect==7.5.0
+iniconfig==2.3.0
+intel-cmplr-lib-ur==2025.3.3
+intel-openmp==2025.3.3
+ipyevents==2.0.4
+ipyfilechooser==0.6.0
+ipykernel==6.17.1
+ipyleaflet==0.20.0
+ipyparallel==8.8.0
+ipython==7.34.0
+ipython-genutils==0.2.0
+ipython-sql==0.5.0
+ipywidgets==7.7.1
+isoduration==20.11.0
+itsdangerous==2.2.0
+jaraco.classes==3.4.0
+jaraco.context==6.1.2
+jaraco.functools==4.5.0
+jax==0.7.2
+jax-cuda12-pjrt==0.7.2
+jax-cuda12-plugin==0.7.2
+jaxlib==0.7.2
+jeepney==0.9.0
+jieba==0.42.1
+Jinja2==3.1.6
+jiter==0.16.0
+joblib==1.5.3
+joserfc==1.7.3
+jsonpatch==1.33
+jsonpickle==4.1.2
+jsonpointer==3.1.1
+jsonschema==4.26.0
+jsonschema-specifications==2025.9.1
+jupyter-console==6.6.3
+jupyter-events==0.12.1
+jupyter-leaflet==0.20.0
+jupyter_client==7.4.9
+jupyter_core==5.9.1
+jupyter_kernel_gateway @ git+https://github.com/googlecolab/kernel_gateway@b134e9945df25c2dcb98ade9129399be10788671
+jupyter_server==2.20.0
+jupyter_server_terminals==0.5.4
+jupyterlab_pygments==0.3.0
+jupyterlab_widgets==3.0.16
+jupytext==1.19.4
+kaggle==2.0.2
+kagglehub==1.0.2
+kagglesdk==0.1.23
+keras==3.13.2
+keras-hub==0.26.0
+keras-nlp==0.26.0
+keyring==25.7.0
+keyrings.google-artifactregistry-auth==1.1.2
+kiwisolver==1.5.0
+langchain==1.3.13
+langchain-core==1.4.9
+langchain-protocol==0.0.18
+langgraph==1.2.9
+langgraph-checkpoint==4.1.1
+langgraph-prebuilt==1.1.0
+langgraph-sdk==0.4.2
+langsmith==0.10.2
+lark==1.3.1
+launchpadlib==1.10.16
+lazr.restfulclient==0.14.4
+lazr.uri==1.0.6
+lazy-loader==0.5
+libclang==18.1.1
+libcudf-cu12==26.2.1
+libcugraph-cu12==26.2.0
+libcuml-cu12==26.2.0
+libcuvs-cu12==26.2.0
+libkvikio-cu12==26.2.0
+libpysal==4.14.1
+libraft-cu12==26.2.0
+librmm-cu12==26.2.0
+librosa==0.11.0
+libucx-cu12==1.19.0
+libucxx-cu12==0.48.0
+lightgbm==4.6.0
+linkify-it-py==2.1.0
+llvmlite==0.43.0
+locket==1.0.0
+logical-unification==0.4.7
+lxml==6.1.1
+Mako==1.1.3
+mapclassify==2.10.0
+Markdown==3.10.2
+markdown-it-py==4.2.0
+MarkupSafe==3.0.3
+matplotlib==3.10.0
+matplotlib-inline==0.2.2
+matplotlib-venn==1.1.2
+mdit-py-plugins==0.6.1
+mdurl==0.1.2
+mgwr==2.2.1
+miniKanren==1.0.5
+missingno==0.5.2
+mistune==3.3.3
+mizani==0.13.5
+mkl==2025.3.1
+ml_dtypes==0.5.4
+mlxtend==0.23.4
+mmh3==5.2.1
+momepy==0.11.0
+more-itertools==10.8.0
+moviepy==1.0.3
+mpmath==1.3.0
+msgpack==1.2.1
+multidict==6.7.1
+multipledispatch==1.0.0
+multiprocess==0.70.16
+multitasking==0.0.13
+murmurhash==1.0.15
+music21==9.9.2
+namex==0.1.0
+narwhals==2.24.0
+natsort==8.4.0
+nbclassic==1.3.3
+nbclient==0.10.4
+nbconvert==7.17.1
+nbformat==5.10.4
+ndindex==1.10.1
+nest-asyncio==1.6.0
+networkx==3.6.1
+nh3==0.3.6
+nibabel==5.4.2
+nltk==3.9.1
+notebook==6.5.7
+notebook_shim==0.2.4
+numba==0.60.0
+numba-cuda==0.22.2
+numexpr==2.14.1
+numpy==2.0.2
+nvidia-cublas-cu12==12.8.4.1
+nvidia-cuda-cccl-cu12==12.9.27
+nvidia-cuda-cupti-cu12==12.8.90
+nvidia-cuda-nvcc-cu12==12.8.93
+nvidia-cuda-nvrtc-cu12==12.8.93
+nvidia-cuda-runtime-cu12==12.8.90
+nvidia-cudnn-cu12==9.19.0.56
+nvidia-cufft-cu12==11.3.3.83
+nvidia-cufile-cu12==1.13.1.3
+nvidia-curand-cu12==10.3.9.90
+nvidia-cusolver-cu12==11.7.3.90
+nvidia-cusparse-cu12==12.5.8.93
+nvidia-cusparselt-cu12==0.7.1
+nvidia-libnvcomp-cu12==5.1.0.21
+nvidia-ml-py==13.610.43
+nvidia-nccl-cu12==2.28.9
+nvidia-nvimgcodec-cu12==0.7.0.11
+nvidia-nvjitlink-cu12==12.8.93
+nvidia-nvshmem-cu12==3.4.5
+nvidia-nvtx-cu12==12.8.90
+nvtx==0.2.15
+nx-cugraph-cu12 @ https://pypi.nvidia.com/nx-cugraph-cu12/nx_cugraph_cu12-26.2.0-py3-none-any.whl
+oauth2client==4.1.3
+oauthlib==3.3.1
+omegaconf==2.3.1
+onemkl-license==2025.3.1
+openai==2.45.0
+opencv-contrib-python==4.13.0.92
+opencv-python==5.0.0.93
+opencv-python-headless==5.0.0.93
+openpyxl==3.1.5
+opentelemetry-api==1.42.1
+opentelemetry-resourcedetector-gcp==1.12.0a0
+opentelemetry-sdk==1.42.1
+opentelemetry-semantic-conventions==0.63b1
+opt_einsum==3.4.0
+optax==0.2.8
+optree==0.19.1
+orbax-checkpoint==0.12.1
+orjson==3.11.9
+ormsgpack==1.12.2
+osqp==1.1.3
+packaging==26.2
+pandas==2.2.2
+pandas-datareader==0.11.1
+pandas-gbq==0.30.0
+pandas-stubs==2.2.2.240909
+pandocfilters==1.5.1
+panel==1.9.3
+panel-material-ui==0.14.0
+param==2.4.1
+parso==0.8.7
+parsy==2.2
+partd==1.4.2
+patsy==1.0.2
+peewee==4.2.2
+peft==0.19.1
+pexpect==4.9.0
+pickleshare==0.7.5
+pillow==11.3.0
+platformdirs==4.10.0
+plotly==5.24.1
+plotnine==0.14.5
+pluggy==1.6.0
+plum-dispatch==2.9.0
+pointpats==2.5.5
+polars==1.35.2
+polars-runtime-32==1.35.2
+pooch==1.9.0
+portpicker==1.5.2
+preshed==3.0.13
+prettytable==3.18.0
+proglog==0.1.12
+progressbar2==4.5.0
+prometheus_client==0.25.0
+promise==2.3
+prompt_toolkit==3.0.52
+propcache==0.5.2
+prophet==1.3.0
+proto-plus==1.28.1
+protobuf==5.29.6
+psutil==5.9.5
+psycopg2==2.9.12
+psygnal==0.15.1
+ptyprocess==0.7.0
+PuLP==3.3.2
+py-cpuinfo==9.0.0
+py4j==0.10.9.9
+pyarrow==18.1.0
+pyasn1==0.6.4
+pyasn1_modules==0.4.2
+pycairo==1.29.0
+pycocotools==2.0.11
+pycparser==3.0
+pycryptodomex==3.23.0
+pydantic==2.13.4
+pydantic_core==2.46.4
+pydata-google-auth==1.9.1
+pydot==4.0.1
+pydotplus==2.0.2
+PyDrive2==1.21.1
+pydub==0.25.1
+pyerfa==2.0.1.5
+pygame==2.6.1
+pygit2==1.19.3
+Pygments==2.20.0
+PyGObject==3.48.2
+pyiceberg==0.11.1
+PyJWT==2.13.0
+pylibcudf-cu12==26.2.1
+pylibcugraph-cu12==26.2.0
+pylibraft-cu12==26.2.0
+pymc==5.28.5
+pynndescent==0.6.0
+pyogrio==0.13.0
+pyomo==6.10.1
+PyOpenGL==3.1.10
+pyOpenSSL==26.3.0
+pyparsing==3.3.2
+pyperclip==1.11.0
+pyproj==3.7.2
+pyroaring==1.1.0
+pysal==25.7
+pyshp==3.1.4
+PySocks==1.7.1
+pyspark==4.0.3
+pytensor==2.38.3
+pytest==8.4.2
+python-apt==0.0.0
+python-box==7.4.1
+python-dateutil==2.9.0.post0
+python-dotenv==1.2.2
+python-fasthtml==0.14.6
+python-json-logger==4.1.0
+python-louvain==0.16
+python-multipart==0.0.32
+python-slugify==8.0.4
+python-snappy==0.7.3
+python-utils==4.0.0
+pytz==2025.2
+pyviz_comms==3.0.6
+PyWavelets==1.9.0
+PyYAML==6.0.3
+pyzmq==26.2.1
+quantecon==0.11.3
+raft-dask-cu12==26.2.0
+rapids-dask-dependency==26.2.0
+rapids-logger==0.2.3
+rasterio==1.5.0
+rasterstats==0.21.0
+ratelim==0.1.6
+referencing==0.37.0
+regex==2025.11.3
+requests==2.32.4
+requests-oauthlib==2.0.0
+requests-toolbelt==1.0.0
+requirements-parser==0.9.0
+rfc3339-validator==0.1.4
+rfc3986-validator==0.1.1
+rfc3987-syntax==1.1.0
+rich==13.9.4
+rmm-cu12==26.2.0
+roman-numerals==4.1.0
+roman-numerals-py==4.1.0
+rpds-py==2026.6.3
+rpy2==3.5.17
+rsa==4.9.1
+rtree==1.4.1
+safehttpx==0.1.7
+safetensors==0.8.0
+scikit-image==0.25.2
+scikit-learn==1.6.1
+scipy==1.16.3
+scooby==0.11.2
+scs==3.2.11
+seaborn==0.13.2
+SecretStorage==3.5.0
+segregation==2.5.4
+semantic-version==2.10.0
+Send2Trash==2.1.0
+sentence-transformers==5.6.0
+sentencepiece==0.2.2
+sentry-sdk==2.65.0
+setuptools==75.2.0
+shap==0.52.0
+shapely==2.1.2
+shellingham==1.5.4
+simple-parsing==0.1.8
+simplejson==4.1.1
+simsimd==6.5.16
+six==1.17.0
+sklearn-compat==0.1.6
+sklearn-pandas==2.2.0
+slicer==0.0.8
+smart_open==8.0.0
+smmap==5.0.3
+sniffio==1.3.1
+snowballstemmer==3.1.1
+sortedcontainers==2.4.0
+soundfile==0.14.0
+soupsieve==2.8.4
+soxr==1.1.0
+spacy==3.8.14
+spacy-legacy==3.0.12
+spacy-loggers==1.0.5
+spaghetti==1.7.6
+spanner-graph-notebook==1.1.10
+spglm==1.1.0
+Sphinx==8.2.3
+sphinxcontrib-applehelp==2.0.0
+sphinxcontrib-devhelp==2.0.0
+sphinxcontrib-htmlhelp==2.1.0
+sphinxcontrib-jsmath==1.0.1
+sphinxcontrib-qthelp==2.0.0
+sphinxcontrib-serializinghtml==2.0.0
+spint==1.1.0
+splot==1.1.7
+spopt==0.7.0
+spreg==1.9.0
+SQLAlchemy==2.0.51
+sqlglot==25.20.2
+sqlparse==0.5.5
+srsly==2.5.3
+stanio==0.5.1
+starlette==1.3.1
+statsmodels==0.14.6
+strictyaml==1.7.3
+stringzilla==4.6.2
+stumpy==1.13.0
+sympy==1.14.0
+tables==3.10.2
+tabulate==0.9.0
+tbb==2022.3.1
+tblib==3.2.2
+tcmlib==1.5.0
+tenacity==9.1.4
+tensorboard==2.20.0
+tensorboard-data-server==0.7.2
+tensorflow==2.20.0
+tensorflow-datasets==4.9.10
+tensorflow-hub==0.16.1
+tensorflow-metadata==1.21.0
+tensorflow-probability==0.25.0
+tensorflow-text==2.20.1
+tensorstore==0.1.84
+termcolor==3.3.0
+terminado==0.18.1
+text-unidecode==1.3
+textblob==0.19.0
+tf-slim==1.1.0
+tf_keras==2.20.0
+thinc==8.3.13
+threadpoolctl==3.6.0
+tifffile==2026.4.11
+tiktoken==0.13.0
+timm==1.0.28
+tinycss2==1.5.1
+tobler==0.14.0
+tokenizers==0.22.2
+toml==0.10.2
+tomlkit==0.14.0
+toolz==0.12.1
+torch==2.11.0+cu128
+torchao==0.10.0
+torchaudio==2.11.0+cu128
+torchcodec==0.11.0+cu128
+torchdata==0.11.0
+torchsummary==1.5.1
+torchtune==0.6.1
+torchvision==0.26.0+cu128
+tornado==6.5.7
+tqdm==4.67.3
+traitlets==5.7.1
+traittypes==0.2.3
+transformers==5.13.1
+treelite==4.7.0
+treescope==0.1.10
+triton==3.6.0
+truststore==0.10.4
+tsfresh==0.21.2
+tweepy==4.17.0
+typeguard==4.5.2
+typer==0.26.8
+types-pytz==2026.2.0.20260518
+types-setuptools==82.0.0.20260518
+typing-inspection==0.4.2
+typing_extensions==4.16.0
+tzdata==2026.3
+tzlocal==5.4.4
+uc-micro-py==2.0.0
+ucxx-cu12==0.48.0
+umap-learn==0.5.12
+umf==1.0.3
+uri-template==1.3.0
+uritemplate==4.2.0
+urllib3==2.5.0
+uuid_utils==0.17.0
+uvicorn==0.51.0
+uvloop==0.22.1
+vega-datasets==0.9.0
+wadllib==1.3.6
+wandb==0.28.0
+wasabi==1.1.3
+watchdog==6.0.0
+watchfiles==1.2.0
+wcwidth==0.8.2
+weasel==1.0.0
+webcolors==25.10.0
+webencodings==0.5.1
+websocket-client==1.9.0
+websockets==15.0.1
+Werkzeug==3.1.8
+wheel==0.47.0
+widgetsnbextension==3.6.10
+wordcloud==1.9.6
+wrapt==2.2.2
+xarray==2025.12.0
+xarray-einstats==0.10.0
+xgboost==3.3.0
+xlrd==2.0.2
+xxhash==3.8.1
+xyzservices==2026.3.0
+yarl==1.24.2
+ydf==0.15.0
+ydf_tf==2.20.0
+yellowbrick==1.5
+yfinance==0.2.66
+zict==3.0.0
+zipp==4.1.0
+zstandard==0.25.0

metadata/experiment_config.json

@@ -0,0 +1,105 @@
+{
+  "experiment_id": "cnn-001-participant-disjoint",
+  "created_at_utc": "2026-08-10T04:55:21.052252+00:00",
+  "dataset_repo": "hnam25/asl-hand-gesture-images",
+  "dataset_revision": "8f36ac00ece6dfce94410a980a839d93a912d366",
+  "processed_archive": "ASL_HG_36000/ASL_Processed_Images.zip",
+  "processed_archive_sha256": "a8e7a38c4085fd9dc18aa4fa8646ad7d6917e9ff7374a3ffd5f9f3b37a5c045b",
+  "audit_manifest": {
+    "schema_version": 1,
+    "kind": "audit-only",
+    "created_at_utc": "2026-08-10T03:59:36.546025+00:00",
+    "dataset_repo": "hnam25/asl-hand-gesture-images",
+    "dataset_revision": "bad9dd9297697ec2da901562b203a5a14ebb93d1",
+    "raw_archive": "ASL_HG_36000/ASL_Raw_Images.zip",
+    "raw_archive_sha256": "594cfa0158044085ed61351315c187a6f3a3f9087795b4f4cdba68ecd04f12b1",
+    "classes": [
+      "0",
+      "1",
+      "2",
+      "3",
+      "4",
+      "5",
+      "6",
+      "7",
+      "8",
+      "9",
+      "A",
+      "B",
+      "C",
+      "D",
+      "E",
+      "F",
+      "G",
+      "H",
+      "I",
+      "J",
+      "K",
+      "L",
+      "M",
+      "N",
+      "O",
+      "P",
+      "Q",
+      "R",
+      "S",
+      "T",
+      "U",
+      "V",
+      "W",
+      "X",
+      "Y",
+      "Z"
+    ],
+    "image_extensions": [
+      ".bmp",
+      ".jpeg",
+      ".jpg",
+      ".png",
+      ".webp"
+    ],
+    "totals": {
+      "records": 36000,
+      "readable": 36000,
+      "unreadable": 0,
+      "duplicate_rows": 994,
+      "duplicate_hash_groups": 435
+    }
+  },
+  "split_manifest": {
+    "experiment_id": "cnn-001-participant-disjoint",
+    "policy": "participant-disjoint 8/1/1; one canonical representative per exact raw-image SHA-256",
+    "seed": 42,
+    "participants": {
+      "train": [
+        "P1",
+        "P10",
+        "P3",
+        "P4",
+        "P5",
+        "P6",
+        "P7",
+        "P8"
+      ],
+      "validation": "P2",
+      "test": "P9"
+    },
+    "counts": {
+      "before_deduplication": 36000,
+      "after_deduplication": 35441,
+      "removed_exact_duplicates": 559,
+      "train": 28365,
+      "validation": 3487,
+      "test": 3589
+    }
+  },
+  "model": {
+    "architecture": "CNN from scratch: 3 x [Conv-BN-ReLU-Conv-BN-ReLU-MaxPool]",
+    "image_size": 128,
+    "batch_size": 64,
+    "epochs_max": 30,
+    "learning_rate": 0.001,
+    "dropout": 0.3
+  },
+  "tensorflow": "2.20.0"
+}

metadata/split_manifest.json

@@ -0,0 +1,27 @@
+{
+  "experiment_id": "cnn-001-participant-disjoint",
+  "policy": "participant-disjoint 8/1/1; one canonical representative per exact raw-image SHA-256",
+  "seed": 42,
+  "participants": {
+    "train": [
+      "P1",
+      "P10",
+      "P3",
+      "P4",
+      "P5",
+      "P6",
+      "P7",
+      "P8"
+    ],
+    "validation": "P2",
+    "test": "P9"
+  },
+  "counts": {
+    "before_deduplication": 36000,
+    "after_deduplication": 35441,
+    "removed_exact_duplicates": 559,
+    "train": 28365,
+    "validation": 3487,
+    "test": 3589
+  }
+}

metrics/classification_report.csv

@@ -0,0 +1,40 @@
+,precision,recall,f1-score,support
+0,0.0,0.0,0.0,100.0
+1,1.0,1.0,1.0,100.0
+2,1.0,1.0,1.0,100.0
+3,0.9009009009009009,1.0,0.9478672985781991,100.0
+4,1.0,0.89,0.9417989417989417,100.0
+5,1.0,1.0,1.0,99.0
+6,0.0,0.0,0.0,100.0
+7,0.6216216216216216,0.92,0.7419354838709677,100.0
+8,0.9259259259259259,1.0,0.9615384615384616,100.0
+9,1.0,0.44,0.6111111111111112,100.0
+A,1.0,1.0,1.0,94.0
+B,1.0,0.8,0.8888888888888888,100.0
+C,1.0,1.0,1.0,99.0
+D,1.0,1.0,1.0,100.0
+E,0.5050505050505051,1.0,0.6711409395973155,100.0
+F,1.0,1.0,1.0,100.0
+G,0.0,0.0,0.0,100.0
+H,0.3333333333333333,1.0,0.5,100.0
+I,1.0,1.0,1.0,100.0
+J,1.0,1.0,1.0,100.0
+K,0.8305084745762712,1.0,0.9074074074074074,98.0
+L,1.0,1.0,1.0,100.0
+M,0.047619047619047616,0.05,0.04878048780487805,100.0
+N,0.0,0.0,0.0,100.0
+O,1.0,1.0,1.0,100.0
+P,1.0,1.0,1.0,100.0
+Q,1.0,1.0,1.0,100.0
+R,1.0,1.0,1.0,99.0
+S,1.0,0.97,0.9847715736040609,100.0
+T,1.0,1.0,1.0,100.0
+U,1.0,1.0,1.0,100.0
+V,1.0,1.0,1.0,100.0
+W,0.4791666666666667,0.92,0.6301369863013698,100.0
+X,1.0,1.0,1.0,100.0
+Y,1.0,1.0,1.0,100.0
+Z,1.0,1.0,1.0,100.0
+accuracy,0.8325438840902758,0.8325438840902758,0.8325438840902758,0.8325438840902758
+macro avg,0.7956701798803966,0.8330555555555557,0.800982710569489,3589.0
+weighted avg,0.7951383757649135,0.8325438840902758,0.8004243363709516,3589.0

metrics/confusion_matrix.csv

@@ -0,0 +1,37 @@
+,0,1,2,3,4,5,6,7,8,9,A,B,C,D,E,F,G,H,I,J,K,L,M,N,O,P,Q,R,S,T,U,V,W,X,Y,Z
+0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+1,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+2,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+3,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+4,0,0,0,11,89,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+5,0,0,0,0,0,99,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+6,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0
+7,0,0,0,0,0,0,0,92,8,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+8,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+9,0,0,0,0,0,0,0,56,0,44,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+A,0,0,0,0,0,0,0,0,0,0,94,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+B,0,0,0,0,0,0,0,0,0,0,0,80,0,0,0,0,0,0,0,0,20,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+C,0,0,0,0,0,0,0,0,0,0,0,0,99,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+D,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+E,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+F,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+G,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+H,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+I,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+J,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+K,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,98,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+L,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+M,0,0,0,0,0,0,0,0,0,0,0,0,0,0,95,0,0,0,0,0,0,0,5,0,0,0,0,0,0,0,0,0,0,0,0,0
+N,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0,0,0
+O,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0,0
+P,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0,0
+Q,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0,0,0,0
+R,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,99,0,0,0,0,0,0,0,0
+S,0,0,0,0,0,0,0,0,0,0,0,0,0,0,3,0,0,0,0,0,0,0,0,0,0,0,0,0,97,0,0,0,0,0,0,0
+T,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0,0
+U,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0,0
+V,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0,0,0
+W,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,92,0,0,0
+X,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0,0
+Y,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100,0
+Z,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,100

metrics/summary.json

@@ -0,0 +1,13 @@
+{
+  "experiment_id": "cnn-001-participant-disjoint",
+  "test_accuracy": 0.8325438840902758,
+  "macro_precision": 0.7956701798803966,
+  "macro_recall": 0.8330555555555557,
+  "macro_f1": 0.800982710569489,
+  "O_recall": 1.0,
+  "0_recall": 0.0,
+  "O_to_0": 0,
+  "0_to_O": 0,
+  "best_validation_accuracy": 0.7209635972976685,
+  "epochs_ran": 15
+}

models/cnn_001_participant_disjoint.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7060712e0ac8574f0cbbf0f89f0e5dbc1571c5d5753c2fde6d7e51bd230e8be9
+size 4083344

models/model_summary.txt

@@ -0,0 +1,75 @@
+Model: "cnn_asl"
+┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
+┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
+┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
+│ image (InputLayer)              │ (None, 128, 128, 3)    │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ rescale (Rescaling)             │ (None, 128, 128, 3)    │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ conv2d_6 (Conv2D)               │ (None, 128, 128, 32)   │           864 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ batch_normalization_6           │ (None, 128, 128, 32)   │           128 │
+│ (BatchNormalization)            │                        │               │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ re_lu_6 (ReLU)                  │ (None, 128, 128, 32)   │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ conv2d_7 (Conv2D)               │ (None, 128, 128, 32)   │         9,216 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ batch_normalization_7           │ (None, 128, 128, 32)   │           128 │
+│ (BatchNormalization)            │                        │               │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ re_lu_7 (ReLU)                  │ (None, 128, 128, 32)   │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ max_pooling2d_3 (MaxPooling2D)  │ (None, 64, 64, 32)     │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ dropout_3 (Dropout)             │ (None, 64, 64, 32)     │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ conv2d_8 (Conv2D)               │ (None, 64, 64, 64)     │        18,432 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ batch_normalization_8           │ (None, 64, 64, 64)     │           256 │
+│ (BatchNormalization)            │                        │               │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ re_lu_8 (ReLU)                  │ (None, 64, 64, 64)     │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ conv2d_9 (Conv2D)               │ (None, 64, 64, 64)     │        36,864 │
+├───────────────────────────��─────┼────────────────────────┼───────────────┤
+│ batch_normalization_9           │ (None, 64, 64, 64)     │           256 │
+│ (BatchNormalization)            │                        │               │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ re_lu_9 (ReLU)                  │ (None, 64, 64, 64)     │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ max_pooling2d_4 (MaxPooling2D)  │ (None, 32, 32, 64)     │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ dropout_4 (Dropout)             │ (None, 32, 32, 64)     │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ conv2d_10 (Conv2D)              │ (None, 32, 32, 128)    │        73,728 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ batch_normalization_10          │ (None, 32, 32, 128)    │           512 │
+│ (BatchNormalization)            │                        │               │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ re_lu_10 (ReLU)                 │ (None, 32, 32, 128)    │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ conv2d_11 (Conv2D)              │ (None, 32, 32, 128)    │       147,456 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ batch_normalization_11          │ (None, 32, 32, 128)    │           512 │
+│ (BatchNormalization)            │                        │               │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ re_lu_11 (ReLU)                 │ (None, 32, 32, 128)    │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ max_pooling2d_5 (MaxPooling2D)  │ (None, 16, 16, 128)    │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ dropout_5 (Dropout)             │ (None, 16, 16, 128)    │             0 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ global_average_pooling          │ (None, 128)            │             0 │
+│ (GlobalAveragePooling2D)        │                        │               │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ features (Dense)                │ (None, 256)            │        33,024 │
+├─────────────────────────────────┼────────────────────────┼───────────────┤
+│ dropout (Dropout)               │ (None, 256)            │             0 │
+├─────────────────────────────────┼─────��──────────────────┼───────────────┤
+│ classification (Dense)          │ (None, 36)             │         9,252 │
+└─────────────────────────────────┴────────────────────────┴───────────────┘
+ Total params: 330,628 (1.26 MB)
+ Trainable params: 329,732 (1.26 MB)
+ Non-trainable params: 896 (3.50 KB)
+

reproducibility/08_cnn_baseline_reproducible.ipynb

@@ -0,0 +1,208 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# CNN baseline — ASL-HG, participant-disjoint\n",
+    "\n",
+    "Baseline tái lập: archive processed của ASL-HG, hash metadata đã publish, khử exact duplicate và split theo người tham gia (8/1/1). Không dùng split train/test dựng sẵn của tác giả vì protocol này khóa validation và test theo participant.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%pip -q install 'huggingface-hub>=0.25' pandas pyarrow scikit-learn matplotlib seaborn\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from pathlib import Path\n",
+    "from datetime import datetime, timezone\n",
+    "import hashlib, json, os, random, re, shutil, zipfile\n",
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "import pandas as pd\n",
+    "import seaborn as sns\n",
+    "import tensorflow as tf\n",
+    "from huggingface_hub import snapshot_download\n",
+    "from sklearn.metrics import accuracy_score, classification_report, confusion_matrix\n",
+    "\n",
+    "EXPERIMENT_ID = 'cnn-001-participant-disjoint'\n",
+    "DATASET_REPO = 'hnam25/asl-hand-gesture-images'\n",
+    "DATASET_REVISION = '8f36ac00ece6dfce94410a980a839d93a912d366'\n",
+    "AUDIT_DIRECTORY = 'metadata/colab-audit-2026-08-10'\n",
+    "PROCESSED_ARCHIVE = 'ASL_HG_36000/ASL_Processed_Images.zip'\n",
+    "SEED, IMAGE_SIZE, BATCH_SIZE, EPOCHS, LEARNING_RATE, DROPOUT = 42, 128, 64, 30, 1e-3, .3\n",
+    "CLASSES = [str(i) for i in range(10)] + [chr(i) for i in range(ord('A'), ord('Z') + 1)]\n",
+    "ROOT = Path('/content/asl-cnn-baseline')\n",
+    "HF_ROOT, PROCESSED, OUTPUTS = ROOT/'hf', ROOT/'processed', ROOT/'outputs'\n",
+    "for directory in (HF_ROOT, PROCESSED, OUTPUTS/'models', OUTPUTS/'metrics', OUTPUTS/'figures', OUTPUTS/'logs', OUTPUTS/'metadata'):\n",
+    "    directory.mkdir(parents=True, exist_ok=True)\n",
+    "random.seed(SEED); np.random.seed(SEED); tf.keras.utils.set_random_seed(SEED)\n",
+    "print({'tensorflow': tf.__version__, 'gpus': [d.name for d in tf.config.list_physical_devices('GPU')], 'experiment_id': EXPERIMENT_ID})\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def sha256_file(path, chunk_size=8 * 1024 * 1024):\n",
+    "    digest = hashlib.sha256()\n",
+    "    with Path(path).open('rb') as handle:\n",
+    "        for chunk in iter(lambda: handle.read(chunk_size), b''):\n",
+    "            digest.update(chunk)\n",
+    "    return digest.hexdigest()\n",
+    "\n",
+    "snapshot_download(repo_id=DATASET_REPO, repo_type='dataset', revision=DATASET_REVISION, local_dir=HF_ROOT, allow_patterns=[PROCESSED_ARCHIVE, f'{AUDIT_DIRECTORY}/**'])\n",
+    "archive = HF_ROOT / PROCESSED_ARCHIVE\n",
+    "audit_root = HF_ROOT / AUDIT_DIRECTORY\n",
+    "audit_manifest = json.loads((audit_root/'cache_manifest.json').read_text())\n",
+    "if not archive.is_file(): raise RuntimeError(f'Missing {PROCESSED_ARCHIVE}')\n",
+    "with zipfile.ZipFile(archive) as z:\n",
+    "    for member in z.infolist():\n",
+    "        target = (PROCESSED/member.filename).resolve()\n",
+    "        if PROCESSED.resolve() not in target.parents and target != PROCESSED.resolve(): raise RuntimeError(f'Unsafe ZIP member: {member.filename}')\n",
+    "    z.extractall(PROCESSED)\n",
+    "processed_archive_sha256 = sha256_file(archive)\n",
+    "audit = pd.read_csv(audit_root/'audit.csv')\n",
+    "print({'processed_archive_sha256': processed_archive_sha256, 'audit_totals': audit_manifest['totals']})\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Match every processed image to audited raw-image hash, then retain one representative per exact hash.\n",
+    "image_paths = sorted(path for path in PROCESSED.rglob('*') if path.suffix.lower() in {'.jpg', '.jpeg', '.png', '.bmp', '.webp'})\n",
+    "records = []\n",
+    "subject_pattern = re.compile(r'^P(\\d+)_')\n",
+    "for path in image_paths:\n",
+    "    parts = path.parts\n",
+    "    label = next((part for part in reversed(parts[:-1]) if part in CLASSES), None)\n",
+    "    if label is None: raise RuntimeError(f'Cannot infer label from {path}')\n",
+    "    match = subject_pattern.match(path.name)\n",
+    "    if not match: raise RuntimeError(f'Cannot infer participant from {path.name}')\n",
+    "    records.append({'label': label, 'processed_path': str(path), 'relative_path': f'{label}/{path.name}', 'participant': f'P{match.group(1)}'})\n",
+    "processed = pd.DataFrame(records)\n",
+    "usable = processed.merge(audit[audit.status == 'ok'][['relative_path', 'label', 'sha256']], on=['relative_path', 'label'], how='left', validate='one_to_one')\n",
+    "if len(usable) != 36000 or usable.sha256.isna().any(): raise RuntimeError('Processed archive does not match the pinned audit metadata.')\n",
+    "if (usable.groupby('sha256').label.nunique() > 1).any(): raise RuntimeError('Conflicting labels for an exact raw-image hash.')\n",
+    "usable = usable.sort_values(['sha256', 'relative_path'], kind='stable').reset_index(drop=True)\n",
+    "usable['duplicate_group_size'] = usable.groupby('sha256').sha256.transform('size')\n",
+    "usable['canonical_relative_path'] = usable.groupby('sha256').relative_path.transform('first')\n",
+    "usable['is_canonical'] = usable.relative_path.eq(usable.canonical_relative_path)\n",
+    "usable.to_csv(OUTPUTS/'metadata'/'deduplication_manifest.csv', index=False)\n",
+    "before_dedup = len(usable); usable = usable[usable.is_canonical].copy()\n",
+    "participants = sorted(usable.participant.unique(), key=lambda value: int(value[1:]))\n",
+    "if len(participants) != 10 or set(usable.participant) != set(participants): raise RuntimeError(f'Expected P1-P10, got {participants}')\n",
+    "rng = np.random.default_rng(SEED); ordered = list(rng.permutation(participants))\n",
+    "train_participants, validation_participant, test_participant = sorted(ordered[:8]), ordered[8], ordered[9]\n",
+    "train = usable[usable.participant.isin(train_participants)].copy()\n",
+    "validation = usable[usable.participant.eq(validation_participant)].copy()\n",
+    "test = usable[usable.participant.eq(test_participant)].copy()\n",
+    "for name, frame in {'train': train, 'validation': validation, 'test': test}.items(): frame[['processed_path', 'label', 'participant', 'sha256']].to_csv(OUTPUTS/'metadata'/f'{name}.csv', index=False)\n",
+    "if set(train.sha256) & set(validation.sha256) or set(train.sha256) & set(test.sha256) or set(validation.sha256) & set(test.sha256): raise RuntimeError('Hash leakage detected.')\n",
+    "if set(train.participant) & set(validation.participant) or set(train.participant) & set(test.participant) or set(validation.participant) & set(test.participant): raise RuntimeError('Participant leakage detected.')\n",
+    "split_manifest = {'experiment_id': EXPERIMENT_ID, 'policy': 'participant-disjoint 8/1/1; one canonical representative per exact raw-image SHA-256', 'seed': SEED, 'participants': {'train': train_participants, 'validation': validation_participant, 'test': test_participant}, 'counts': {'before_deduplication': before_dedup, 'after_deduplication': len(usable), 'removed_exact_duplicates': before_dedup-len(usable), 'train': len(train), 'validation': len(validation), 'test': len(test)}}\n",
+    "(OUTPUTS/'metadata'/'split_manifest.json').write_text(json.dumps(split_manifest, indent=2), encoding='utf-8')\n",
+    "print(json.dumps(split_manifest, indent=2))\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "label_index = {label: index for index, label in enumerate(CLASSES)}\n",
+    "def make_dataset(frame, training=False):\n",
+    "    ds = tf.data.Dataset.from_tensor_slices((frame.processed_path.values, frame.label.map(label_index).values))\n",
+    "    if training: ds = ds.shuffle(len(frame), seed=SEED, reshuffle_each_iteration=True)\n",
+    "    def load(path, label):\n",
+    "        image = tf.io.decode_image(tf.io.read_file(path), channels=3, expand_animations=False); image.set_shape([None, None, 3])\n",
+    "        image = tf.image.resize(tf.cast(image, tf.float32), (IMAGE_SIZE, IMAGE_SIZE))\n",
+    "        return image, tf.one_hot(label, len(CLASSES))\n",
+    "    return ds.map(load, num_parallel_calls=tf.data.AUTOTUNE).batch(BATCH_SIZE).prefetch(tf.data.AUTOTUNE)\n",
+    "\n",
+    "train_ds, validation_ds, test_ds = make_dataset(train, True), make_dataset(validation), make_dataset(test)\n",
+    "def build_cnn():\n",
+    "    inputs = tf.keras.Input((IMAGE_SIZE, IMAGE_SIZE, 3), name='image')\n",
+    "    x = tf.keras.layers.Rescaling(1/255, name='rescale')(inputs)\n",
+    "    for filters in (32, 64, 128):\n",
+    "        x = tf.keras.layers.Conv2D(filters, 3, padding='same', use_bias=False)(x)\n",
+    "        x = tf.keras.layers.BatchNormalization()(x); x = tf.keras.layers.ReLU()(x)\n",
+    "        x = tf.keras.layers.Conv2D(filters, 3, padding='same', use_bias=False)(x)\n",
+    "        x = tf.keras.layers.BatchNormalization()(x); x = tf.keras.layers.ReLU()(x)\n",
+    "        x = tf.keras.layers.MaxPooling2D()(x); x = tf.keras.layers.Dropout(DROPOUT / 2)(x)\n",
+    "    x = tf.keras.layers.GlobalAveragePooling2D(name='global_average_pooling')(x)\n",
+    "    x = tf.keras.layers.Dense(256, activation='relu', name='features')(x)\n",
+    "    x = tf.keras.layers.Dropout(DROPOUT, name='dropout')(x)\n",
+    "    outputs = tf.keras.layers.Dense(len(CLASSES), activation='softmax', name='classification')(x)\n",
+    "    return tf.keras.Model(inputs, outputs, name='cnn_asl')\n",
+    "\n",
+    "model = build_cnn(); model.compile(optimizer=tf.keras.optimizers.Adam(LEARNING_RATE), loss='categorical_crossentropy', metrics=['accuracy'])\n",
+    "with (OUTPUTS/'models'/'model_summary.txt').open('w') as handle: model.summary(print_fn=lambda line: handle.write(line+'\\n'))\n",
+    "checkpoint = OUTPUTS/'models'/'cnn_001_participant_disjoint.keras'\n",
+    "callbacks = [tf.keras.callbacks.ModelCheckpoint(checkpoint, monitor='val_accuracy', save_best_only=True), tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=5, restore_best_weights=True), tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss', patience=2, factor=.2), tf.keras.callbacks.CSVLogger(OUTPUTS/'logs'/'training_history.csv')]\n",
+    "history = model.fit(train_ds, validation_data=validation_ds, epochs=EPOCHS, verbose=2, callbacks=callbacks)\n",
+    "pd.DataFrame(history.history).to_csv(OUTPUTS/'logs'/'training_history.csv', index=False)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "best = tf.keras.models.load_model(checkpoint)\n",
+    "probability = best.predict(test_ds, verbose=1); predicted = probability.argmax(1); truth = test.label.map(label_index).to_numpy()\n",
+    "report = classification_report(truth, predicted, labels=range(36), target_names=CLASSES, output_dict=True, zero_division=0)\n",
+    "matrix = confusion_matrix(truth, predicted, labels=range(36))\n",
+    "pd.DataFrame(report).T.to_csv(OUTPUTS/'metrics'/'classification_report.csv')\n",
+    "pd.DataFrame(matrix, index=CLASSES, columns=CLASSES).to_csv(OUTPUTS/'metrics'/'confusion_matrix.csv')\n",
+    "o, zero = label_index['O'], label_index['0']\n",
+    "summary = {'experiment_id': EXPERIMENT_ID, 'test_accuracy': float(accuracy_score(truth, predicted)), 'macro_precision': report['macro avg']['precision'], 'macro_recall': report['macro avg']['recall'], 'macro_f1': report['macro avg']['f1-score'], 'O_recall': report['O']['recall'], '0_recall': report['0']['recall'], 'O_to_0': int(matrix[o, zero]), '0_to_O': int(matrix[zero, o]), 'best_validation_accuracy': float(max(history.history['val_accuracy'])), 'epochs_ran': len(history.history['loss'])}\n",
+    "(OUTPUTS/'metrics'/'summary.json').write_text(json.dumps(summary, indent=2), encoding='utf-8')\n",
+    "plt.figure(figsize=(16, 13)); sns.heatmap(matrix, cmap='Blues', xticklabels=CLASSES, yticklabels=CLASSES); plt.xlabel('Predicted'); plt.ylabel('True'); plt.tight_layout(); plt.savefig(OUTPUTS/'figures'/'confusion_matrix.png', dpi=180); plt.close()\n",
+    "config = {'experiment_id': EXPERIMENT_ID, 'created_at_utc': datetime.now(timezone.utc).isoformat(), 'dataset_repo': DATASET_REPO, 'dataset_revision': DATASET_REVISION, 'processed_archive': PROCESSED_ARCHIVE, 'processed_archive_sha256': processed_archive_sha256, 'audit_manifest': audit_manifest, 'split_manifest': split_manifest, 'model': {'architecture': 'CNN from scratch: 3 x [Conv-BN-ReLU-Conv-BN-ReLU-MaxPool]', 'image_size': IMAGE_SIZE, 'batch_size': BATCH_SIZE, 'epochs_max': EPOCHS, 'learning_rate': LEARNING_RATE, 'dropout': DROPOUT}, 'tensorflow': tf.__version__}\n",
+    "(OUTPUTS/'metadata'/'experiment_config.json').write_text(json.dumps(config, indent=2), encoding='utf-8')\n",
+    "os.system(f\"pip freeze > {OUTPUTS/'metadata'/'environment.txt'}\")\n",
+    "print(json.dumps(summary, indent=2))\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Publish\n",
+    "\n",
+    "Download `outputs/` về local. Tạo một Hugging Face model repo và upload nguyên thư mục output cùng notebook này; model card phải nêu rõ participant-disjoint protocol, revision dữ liệu, SHA archive và số liệu test.\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "name": "python",
+   "version": "3.x"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

reproducibility/requirements.txt

@@ -0,0 +1,13 @@
+tensorflow>=2.16,<2.20
+mediapipe>=0.10.14
+opencv-python>=4.9
+numpy>=1.26
+pandas>=2.2
+scikit-learn>=1.5
+matplotlib>=3.9
+seaborn>=0.13
+PyYAML>=6.0
+pyarrow>=16.0
+jupyterlab>=4.0
+pytest>=8.0
+huggingface-hub>=0.25