Home Explore Help
Sign In
junxiaoyao
/
RT-ODLab
1
0
Fork 0
Files Issues 0 Pull Requests 0 Wiki
Tree: 399ec4dafb
Branches Tags
RT-ODLab
/
models
/
detectors
/
rtcdet
yjh0410 2bcfbab4ed release RTCDet-S on COCO 2 years ago
..
README.md 2bcfbab4ed release RTCDet-S on COCO 2 years ago
build.py 578dc295f4 redesign RTCDet 2 years ago
loss.py bc2f5d6362 train RTCDet-S 2 years ago
matcher.py f6753b0e9b train RTCDet-S 2 years ago
rtcdet.py f6753b0e9b train RTCDet-S 2 years ago
rtcdet_backbone.py 9650add52f fix a bug in YoloBottleNeck 2 years ago
rtcdet_basic.py 9650add52f fix a bug in YoloBottleNeck 2 years ago
rtcdet_head.py 578dc295f4 redesign RTCDet 2 years ago
rtcdet_neck.py 578dc295f4 redesign RTCDet 2 years ago
rtcdet_pafpn.py 62a5b79e6a update 2 years ago
rtcdet_pred.py f6753b0e9b train RTCDet-S 2 years ago

README.md

RTCDet-v2: My Second Empirical Study of Real-Time Convolutional Object Detectors.

Model Scale Batch APtest
0.5:0.95 		APtest
0.5 		APval
0.5:0.95 		APval
0.5 		FLOPs
(G) 		Params
(M) 		Weight
RTCDet-N 640 8xb16
RTCDet-T 640 8xb16
RTCDet-S 640 8xb16 44.7 63.7 31.5 8.4 ckpt
RTCDet-M 640 8xb16
RTCDet-L 640 8xb16
RTCDet-X 640
  • For training, we train my RTCDet series series with 300 epochs on COCO.
  • For data augmentation, we use the large scale jitter (LSJ), Mosaic augmentation and Mixup augmentation, following the setting of YOLOX, but we remove the rotation transformation which is used in YOLOX's strong augmentation.
  • For optimizer, we use AdamW with weight decay 0.05 and base per image lr 0.001 / 64.
  • For learning rate scheduler, we use linear decay scheduler.
  • Due to my limited computing resources, I can not train RTCDet-X with the setting of batch size=128.
Model Scale Batch APtest
0.5:0.95 		APtest
0.5 		APval
0.5:0.95 		APval
0.5 		FLOPs
(G) 		Params
(M) 		Weight
RTCDet-P 320 8xb16
RTCDet-P 416 8xb16
RTCDet-P 512 8xb16
RTCDet-P 640 8xb16