Evaluating UniPixel

🛠️ Environment Setup

Please refer to TRAIN.md for setting up the environment.

📚 Checkpoint Preparation

Download the checkpoints from Hugging Face and place them into the model_zoo folder.

UniPixel
└─ model_zoo
   ├─ UniPixel-3B
   └─ UniPixel-7B

📦 Dataset Preparation

Download the desired datasets / benchmarks from Hugging Face, extract them, and place them into the data folder. The processed files should be organized in the following structure (taking ref_youtube_vos as an example).

UniPixel
└─ data
   └─ ref_youtube_vos
      ├─ meta_expressions
      ├─ train
      ├─ valid
      └─ mask_dict.pkl

🔮 Start Evaluation

Use the following command to evaluate UniPixel automatically on all benchmarks. The default setting is to distribute the samples to multiple GPUs/NPUs for acceleration.

bash scripts/auto_eval.sh <path-to-checkpoint>

You may comment out some datasets in auto_eval.sh if you don't need them.

The inference outputs and evaluation metrics will be saved into the <path-to-checkpoint> folder by default.