Vision modality not enabled in Android app
Created a basic Android app and pushed model to phone, model initialises but cannot accept image input (image also resized correct). Using Mediapipe implementation("com.google.mediapipe:tasks-genai:0.10.29").
Error:
%aUNAVAILABLE: Vision Modality is not enabled
Does this model currently not have image input enabled?
Thanks
Hi @Deeko76 ,
Yes, the model accepts image as an input. Here are the constraints: Images, normalized to 256x256, 512x512, or 768x768 resolution and encoded to 256 tokens each.
This looks like you haven't enabled vision support for LLM Inference API, you have to set EnableVisionModality configuration option to true within the Graph options.
Here is the official reference: https://ai.google.dev/edge/mediapipe/solutions/genai/llm_inference/android#image-input
Thank you!