Trace Model Visualizer

Upload an image and provide a natural language task instruction to predict the trajectory/trace using mihirgrao/trace-model.

The model predicts coordinate points from your instruction; they are overlaid on the image (green → red gradient) and listed below.

Upload Image

Natural language instruction

Enter a task description in natural language. The model predicts the trace for this instruction.

Local model (if no eval server selected)

Model ID

Hugging Face model ID (auto-loads on first inference if no eval server selected)

Use Franka / qwenvl inference

Uses preprocess_qwen_visual (qwenvl). Requires qwen-vl-finetune on PYTHONPATH.

Prompt sent to model:

Predict the trajectory or trace in this image. Output the coordinates as a list of [x, y] pairs, e.g. [[0.1, 0.2], [0.3, 0.4], ...]. Use normalized coordinates between 0 and 1.

Image with Trace Overlay

Model Prediction (raw)

Select an eval server from the sidebar (auto-connects), or run inference with local model.

Trace Model Visualizer

🔧 Model Configuration

📋 Model Information

Local model (if no eval server selected)