Instructions to use ThienDan2705/countyolo-checkpoints with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- sam2
How to use ThienDan2705/countyolo-checkpoints with sam2:
# Use SAM2 with images import torch from sam2.sam2_image_predictor import SAM2ImagePredictor predictor = SAM2ImagePredictor.from_pretrained("ThienDan2705/countyolo-checkpoints") 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("ThienDan2705/countyolo-checkpoints") 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
CountYOLO project checkpoint collection
Public checkpoint storage for https://github.com/Thien-Dan/countyolo (master).
Project-trained CountYOLO density heads
- exnoise: recommended image head based on existing FSC147 evaluations.
- seed3: selected head for the existing video pilot.
- exmix2: comparison head.
These heads need the project architecture and pretrained backbones. Image and video rankings use different configurations.
Upstream checkpoints (unmodified mirrors)
sam3.pt: Meta SAM 3, originally https://huggingface.co/facebook/sam3 . Copyright Meta. Distributed under the SAM License; seelicenses/SAM3-LICENSE.txt. Use and redistribution remain subject to those terms. Acknowledge SAM Materials in research publications.countgd_box.pth: official CountGD-Box checkpoint provided by CountVid (https://github.com/niki-amini-naieni/CountVid), original download https://drive.google.com/file/d/1bw-YIS-Il5efGgUqGVisIZ8ekrhhf_FD/view . Copyright Niki Amini-Naieni; see the preserved CountVid and CountGD MIT license notices inlicenses/.sam2.1_hiera_large.pt: Meta SAM 2.1 large, original https://dl.fbaipublicfiles.com/segment_anything_2/092824/sam2.1_hiera_large.pt . Seelicenses/SAM2-LICENSE.txtand upstream notices (Apache 2.0).
CountVid is a pipeline using CountGD-Box and SAM 2.1, not a single checkpoint. BERT, YOLO-World and DINOv2 backbones are still obtained using their existing upstream setup. This is a project convenience mirror, not an official Meta or CountVid release. Each upstream file retains its original license; no common license replaces these terms.
See checksums.json for paths, upstream sources, byte sizes and SHA-256 checksums. Download using tools/download_checkpoints.py --group all in the code repo, or choose countyolo, sam3, or countvid.
- Downloads last month
- -