yjh0410 fd8e9e5e0c update ELANNet-v2		2 年之前
..
README.md	6f4f753562 keep training YOLOX-L	2 年之前
build.py	2eb02320fc release RTCDet-v1 & rename RTMDet to RTCDet	2 年之前
loss.py	a1b69fb635 update	2 年之前
matcher.py	8d590a1158 modify RTCDet-v2	2 年之前
rtcdet_v2.py	5754e59167 debug RTCDet-v2	2 年之前
rtcdet_v2_backbone.py	fd8e9e5e0c update ELANNet-v2	2 年之前
rtcdet_v2_basic.py	fd8e9e5e0c update ELANNet-v2	2 年之前
rtcdet_v2_head.py	fd8e9e5e0c update ELANNet-v2	2 年之前
rtcdet_v2_neck.py	fd8e9e5e0c update ELANNet-v2	2 年之前
rtcdet_v2_pafpn.py	fd8e9e5e0c update ELANNet-v2	2 年之前
rtcdet_v2_pred.py	2eb02320fc release RTCDet-v1 & rename RTMDet to RTCDet	2 年之前

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

For training, we train my RTCDetv2 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 RTCDetv2-X with the setting of batch size=128.