Instructions to use nkkbr/ViCA2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use nkkbr/ViCA2 with Transformers:
# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("nkkbr/ViCA2", dtype="auto") - sam2
How to use nkkbr/ViCA2 with sam2:
# Use SAM2 with images import torch from sam2.sam2_image_predictor import SAM2ImagePredictor predictor = SAM2ImagePredictor.from_pretrained(nkkbr/ViCA2) with torch.inference_mode(), torch.autocast("cuda", dtype=torch.bfloat16): predictor.set_image(<your_image>) masks, _, _ = predictor.predict(<input_prompts>)# Use SAM2 with videos import torch from sam2.sam2_video_predictor import SAM2VideoPredictor predictor = SAM2VideoPredictor.from_pretrained(nkkbr/ViCA2) with torch.inference_mode(), torch.autocast("cuda", dtype=torch.bfloat16): state = predictor.init_state(<your_video>) # add new prompts and instantly get the output on the same frame frame_idx, object_ids, masks = predictor.add_new_points(state, <your_prompts>): # propagate the prompts to get masklets throughout the video for frame_idx, object_ids, masks in predictor.propagate_in_video(state): ... - Notebooks
- Google Colab
- Kaggle
evaluation different from your self-reported results
Hi, sorry to bother you again. I use your checkpoint nkkbr/vica2 to evaluate VSI-Bench, and the results I got are a bit different from your reported results-60%. The evaluation results I got is 57% and I am wondering what settings might cause this difference (to provide additional information, I used the 64 frames and the temperature 0 during the evaluation and followed the QA template you used on the github)? Thanks for your patience!
Hi, @catherinexyz , thanks for reaching out and for your evaluation of ViCA2 on VSI-Bench.
To be precise, our tested result on VSI-Bench is 56.8%.
You can find a detailed breakdown of our scores across the eight tasks in the main table of our repository:
Overall Performance on VSI-Bench
To help us compare results more granularly, could you please share the scores you obtained for each of the eight tasks in your test?
Thanks again for your valuable feedback!
That makes sense! Thanks for your reply!