Detection models

14 detection models. Each section says what the model is, whether it may be used commercially and how fast it is; licences and checks are under Details.

Input: one RGB image. Output: boxes, class labels, scores and instance masks.

Measured on RTX 5060 Ti, torch 2.14.1+cu130; fastforks bench MODEL --size HxW measures yours. The summary uses one input, 800x1344: kernels time with its speed-up over the faster original run, and kernels peak reserved memory with its ratio to the smaller original one. Where measured (one variant per family), a variant's table also lists portable under torch.compile and the fp16 TensorRT engine of portable (export; memory: the engine's device memory). – = not measured.

Other devices: Apple M1 Pro 16 GB, NVIDIA L4, NVIDIA L40S, NVIDIA A10, NVIDIA A100-SXM4-40GB, NVIDIA RTX PRO 6000; across devices.

All models

Modes, licences, training data and outputs per model

Commercial status per variant: yes if the code and the weights licence both allow commercial use (the training data is listed but does not change it; weights without a licence of their own are assumed to carry the code licence); no if either forbids it; unclear for custom licence terms (treat as no).

model	modes	code licence	weights licence	training data	commercial	outputs
detr-r101-coco	portable, kernels	Apache-2.0	Apache-2.0	coco, imagenet-1k	yes	boxes, scores, labels
detr-r50-coco	portable, kernels	Apache-2.0	Apache-2.0	coco, imagenet-1k	yes	boxes, scores, labels
dfine-l-coco	portable, kernels	Apache-2.0	Apache-2.0	coco, imagenet-1k	yes	boxes, scores, labels
dfine-m-coco	portable, kernels	Apache-2.0	Apache-2.0	coco, imagenet-1k	yes	boxes, scores, labels
dfine-n-coco	portable, kernels	Apache-2.0	Apache-2.0	coco, imagenet-1k	yes	boxes, scores, labels
dfine-s-coco	portable, kernels	Apache-2.0	Apache-2.0	coco, imagenet-1k	yes	boxes, scores, labels
dfine-x-coco	portable, kernels	Apache-2.0	Apache-2.0	coco, imagenet-1k	yes	boxes, scores, labels
faster-rcnn-r50-fpn-3x-coco	portable, kernels	Apache-2.0	CC-BY-SA-3.0	coco	yes	boxes, scores, labels, masks
mask-rcnn-r50-fpn-3x-coco	portable, kernels	Apache-2.0	CC-BY-SA-3.0	coco	yes	boxes, scores, labels, masks
rtdetrv2-l-coco	portable, kernels	Apache-2.0	Apache-2.0 (assumed from code)	coco, imagenet-1k	yes	boxes, scores, labels
rtdetrv2-m-coco	portable, kernels	Apache-2.0	Apache-2.0 (assumed from code)	coco, imagenet-1k	yes	boxes, scores, labels
rtdetrv2-ms-coco	portable, kernels	Apache-2.0	Apache-2.0 (assumed from code)	coco, imagenet-1k	yes	boxes, scores, labels
rtdetrv2-s-coco	portable, kernels	Apache-2.0	Apache-2.0 (assumed from code)	coco, imagenet-1k	yes	boxes, scores, labels
rtdetrv2-x-coco	portable, kernels	Apache-2.0	Apache-2.0 (assumed from code)	coco, imagenet-1k	yes	boxes, scores, labels

Training datasets