Biodiversity

What's New?

Deepfaune in Our Model Zoo!!

• We are excited to announce the release of the Deepfaune models—both the detector and classifier—in PyTorch-Wildlife, adding to our growing model zoo. A huge thank you to the Deepfaune team for your support! Deepfaune is one of the most comprehensive models focused on the European ecosystem for both detection and classification. It serves as a great complement to MegaDetector, which has primarily been trained on datasets from North America, South America, and Africa. The Deepfaune detector is also our first third-party camera trap detection model integrated into PyTorch-Wildlife!
• To use the model, you just need to load them as any other Pytorch-Wildife models:

detection_model = pw_detection.DeepfauneDetector(device=DEVICE)
classification_model = pw_classification.DeepfauneClassifier(device=DEVICE)

• You can also use the detection_classification_pipeline_demo.py script in the demo folder to test the whole detection + classification pipeline.
• Please also take a look at the original Deepfaune website and give them a star!

Deepfaune-New-England in Our Model Zoo Too!!

• Besides the original Deepfaune mode, there is another fine-tuned Deepfaune model developed by USGS for the Northeastern NA area called Deepfaune-New-England (DFNE). It can also be loaded with classification_model = pw_classification.DFNE(device=DEVICE)
• Please take a look at the orignal DFNE repo and give them a star!

SpeciesNet is available in Pytorch-Wildlife for testing!

• We have added SpeciesNet into our model zoo, which is compatible with all detection models provided by Pytorch-Wildlife. Please refer to this document for more details!

What's New

• In this version of Pytorch-Wildlife, we are happy to release our detection fine-tuning module, with which users can fine-tune their own detection model from any released pre-trained MegaDetectorV6 models. Besides, this module also has functionalities that help users to prepare their datasets for the fine-tuning, just as our classification fine-tuning modules. For more details, please check the readme. Currently the fine-tuning is based on Ultralytics with AGPL. We will release MIT versions in the future.

• We have also released additional MegaDetectorV6 models based on Yolo-v10 and RtDetr. We have skipped Yolo-v11 models because of limited performance and architectural gains. In the following figure we can see the Parameter to Performance metrics of our released models. Users can choose according to specific project needs. All of the MegaDetectorV6 models have at least 50% less parameters than MegaDetectorV5 and still have better animal detection performance. Our smallest released model is MegaDetectorV6-yolov10-c, which only has 2% of the parameter size of MegaDetectorV5 but still has 4% better animal detection recall.

• In terms of the MIT and Apache versions, most of them have also finished training but are waiting for internal review before they can be released.

• We have also updated our AI4G-Amazon model with bigger datasets and it has a better performance compared to previous iterations. Please feel free to test it or fine-tune on it.

• We will also make a new roadmap for 2025 in the next couple of updates.

🎉🎉🎉 Pytorch-Wildlife Version 1.1.0

MegaDetectorV6 is finally out

After a few months of public beta testing, we are finally ready to officially release our 6th version of MegaDetector, MegaDetectorV6! In the next generation of MegaDetector, we are focusing on computational efficiency, performance, mordernizing of model architectures, and licensing. We have trained multiple new models using different model architectures, including Yolo-v9, Yolo-v11, and RT-Detr for maximum user flexibility. We have a rolling release schedule for different versions of MegaDetectorV6, and in the first step, we are releasing the compact version of MegaDetectorV6 with Yolo-v9 (MDv6-ultralytics-yolov9-compact, MDv6-c in short). From now on, we encourage our users to use MegaDetectorV6 as their default animal detection model.

This MDv6-c model is optimized for performance and low-budget devices. It has only one-sixth (SMALLER) of the parameters of the previous MegaDetectorV5 and exhibits 12% higher recall (BETTER) on animal detection in our validation datasets. In other words, MDv6-c has significantly fewer false negatives when detecting animals, making it a more robust animal detection model than MegaDetectorV5. Furthermore, one of our testers reported that the speed of MDv6-c is at least 5 times FASTER than MegaDetectorV5 on their datasets.

Learn how to use MegaDetectorV6 in our image demo and video demo.

HerdNet in Pytorch-Wildlife

We have incorporated a point-based overhead animal detection model into our model zoo called HerdNet (Delplanque et al. 2022). Two model weights are incorporated in this release, HerdNet-general (their default weights) and HerdNet-ennedi (their model trained on Ennedi 2019 datasets). More details can be found here and in their original repo. This is the first third-party model in Pytorch-Wildlife and the foundation of our expansion to overhead/aerial animal detection and classification. Please see our HerdNet demo on how to use it!

Customed Weights

You can now load custom weights you fine-tuned on your own datasets using the finetuning module directly in the Pytorch-Wildlife pipeline! Please see the demo on how to do it. You can also load it in our Gradio app!

Folder Separation

You can now automatically separate your image detections into folders based on detection results! Please see our folder separation demo on how to do it. You can also test it in our Gradio demo!

Simplified Batch Detection!

We have also simplified the batch detection pipeline. Now we do not need to define pytorch datasets and dataloaders specifically.

Previous:

tgt_folder_path = os.path.join(".","demo_data","imgs")

dataset = pw_data.DetectionImageFolder(
    tgt_folder_path,
    transform=pw_trans.MegaDetector_v5_Transform(target_size=detection_model.IMAGE_SIZE,
                                                 stride=detection_model.STRIDE),
    extension='JPG' 

loader = DataLoader(dataset, batch_si

Hello everyone! We have just updated Pytorch-Wildlife to version 1.0.2.4. This is a small update but fixes some bugs related to result outputs reported in issue https://github.com/microsoft/CameraTraps/issues/428 . Besides the bug fixes, we have also added a HuggingFace demo and have two google colab demos Image_detection_demo and Video_detection_demo. We also have fully revamped our README for better visual and created a discord server for our community. Please join us! Thank you so much!