-
## 环境/Environment
ubuntu 18.04
paddle-det 2.2.0
paddlepaddle-gpu 2.1.0
python 3.7
CUDA 10.2.89
cuDNN 7
## 问题
目标想要训练(火焰,烟雾,行人)三类的模组,gpu单卡训练, 可是mAP卡在30-40,loss也在20-30,实际测试起来效果没有很好...…
-
问题描述:
1、pytorch训练模型,导出onnx,再转到paddle,paddle模型没有实现yolov5后处理部分(解码+nms),请问怎么实现呢?
-
Hi All,
I am working on research which compare all YOLO detectors accuracy based on thermal, depth and merged images and I wanted include also PP YOLO, but cant get that running with paddlepaddle-gpu…
bielu updated
2 years ago
-
联合PP-YOLOV2与Hrnet模型推理
# [联合部署推理] 使用命令如下:
!python deploy/python/keypoint_det_unite_infer.py --det_model_dir=output_inference/ppyolov2_r50vd_dcn_voc/ \
--keypoint_model_dir=output_inference/hrnet_w32…
-
-
![捕获](https://user-images.githubusercontent.com/40631787/144948284-fd14a358-7da0-4dc0-b6eb-ab3083c0482d.JPG)
-
为什么静态图的ppyolo跑不过静态图的yolov3
ghost updated
2 years ago
-
hi,大家好,非常高兴的告诉大家,百度飞桨论文复现赛第五期已经开始了,本次**论文复现赛**共将有100篇的经典&前沿论文供大家复现,以及新增了**工程落地赛**,详细信息可以参考[AI Studio](https://aistudio.baidu.com/aistudio/competition/detail/126/0/introduction),大家是否已经迫不及待了呢~
为了帮助…
-
how to get pp-yolo-v2 train log? I meet some error in train and consider compare log to official log.
-
你好,我想问下,针对前面的各种图像数据处理,不需要损失函数进行约束吗