MM-CamObj

(ARXIV24) This is the official code repository for "MM-CamObj: A Comprehensive Multimodal Dataset for Camouflaged Object Scenarios."

Dataset

• Collected by Fan et al.: SINet-V2
  • COD10K: Download COD10K
• PlantCamo: Download PlantCamo
• CPD1K: Download CPD1K
• Annotated and generated by Cheng et al.: SLT-Net
  • MoCA-Mask: Download MoCA-Mask

To use the dataset, download the files and place them in the dataset/data/ directory. Then, run dataset/data2json.py to convert the data into JSON format.

Evaluation

Make sure to update your API keys and dataset paths accordingly.

1. Modify the dataset and model paths in MM-CamObj/camobjbench/run_eval_batch.sh to match your local setup.
2. Execute the following command to start the evaluation:

   bash MM-CamObj/camobjbench/run_eval_batch.sh

3. For evaluating with GPT Gemini, run the Python scripts located in the following directories:
   • MM-CamObj/camobjbench/eval_gemini-1.5-pro/questions
   • MM-CamObj/camobjbench/eval_GPT4o/questions
   • MM-CamObj/camobjbench/eval_GPT4o_mini/questions

Acknowledgement

Thanks to the open-source code from Mantis