Detic : Detecting Twenty-thousand Classes using Image-level Supervision
Summary :
- Detic trains the classifiers of a detector on image classification data and thus expands the vocabulary of detectors to tens of thousands of concepts. Unlike prior work, Detic does not assign image labels to boxes based on model predictions, making it much easier to implement and compatible with a range of detection architectures and backbones.
- Results show that Detic yields excellent detectors even for classes without box annotations. It outperforms prior work on both open-vocabulary and long-tail detection benchmarks
Features:
- A Detector with image classes that can use image-level labels to easily train detectors.
- Train object detector on image data without box annotation
- Detects any class given class names (using CLIP).
- Cross-dataset generalization to OpenImages and Objects365 without finetuning.
- For the first time, they train a detector with all the twenty-one-thousand classes of the ImageNet dataset and show that it generalizes to new datasets without fine-tuning
- State-of-the-art results on Open-vocabulary LVIS and Open-vocabulary COCO.
- Works for DETR-style detectors.
References :
- arXiv paper : https://lnkd.in/g4k4hyXh
- Code is available at https://lnkd.in/gAENgMhg.
- Colab Link : https://colab.research.google.com/drive/1QtTW9-ukX2HKZGvt0QvVGqjuqEykoZKI
Thanks.
