Facebook Detr Resnet50

The DETR model is an encoder-decoder transformer with a convolutional backbone. Two heads are added on top of the decoder outputs in order to perform object detection: a linear layer for the class labels and a MLP (multi-layer perceptron) for the bounding boxes. The model uses so-called object queries to detect objects in an image. Each object query looks for a particular object in the image. For COCO, the number of object queries is set to 100.

The model is trained using a "bipartite matching loss": one compares the predicted classes + bounding boxes of each of the N = 100 object queries to the ground truth annotations, padded up to the same length N (so if an image only contains 4 objects, 96 annotations will just have a "no object" as class and "no bounding box" as bounding box). The Hungarian matching algorithm is used to create an optimal one-to-one mapping between each of the N queries and each of the N annotations. Next, standard cross-entropy (for the classes) and a linear combination of the L1 and generalized IoU loss (for the bounding boxes) are used to optimize the parameters of the model.

@article{DBLP:journals/corr/abs-2005-12872,
 author    = {Nicolas Carion and
              Francisco Massa and
              Gabriel Synnaeve and
              Nicolas Usunier and
              Alexander Kirillov and
              Sergey Zagoruyko},
 title     = {End-to-End Object Detection with Transformers},
 journal   = {CoRR},
 volume    = {abs/2005.12872},
 year      = {2020},
 url       = {https://arxiv.org/abs/2005.12872},
 archivePrefix = {arXiv},
 eprint    = {2005.12872},
 timestamp = {Thu, 28 May 2020 17:38:09 +0200},
 biburl    = {https://dblp.org/rec/journals/corr/abs-2005-12872.bib},
 bibsource = {dblp computer science bibliography, https://dblp.org}
}