add log parsers for detection models

[zzhfz]

Description

Add automated log parsing scripts for FasterRCNN/YOLOv5/SSD models

Usage Example

1.FasterRCNN

(venv_fasterrcnn) sunjinge@ares1:~/PyTorchModels/Detection/fasterrcnn$ python fasterrcnn_log_parser.py -l fasterrcnn_train.log -o fasterrcnn_metrics.csv
✅ 解析完成！结果已保存到 fasterrcnn_metrics.csv
共处理 13 个epoch，最终mAP: 0.6970
(venv_fasterrcnn) sunjinge@ares1:~/PyTorchModels/Detection/fasterrcnn$ cat fasterrcnn_metrics.csv
epoch,learning_rate,val_mAP,total_loss,throughput(it/s)
1,0.0010,0.4385,0.9696,8.0200
2,0.0010,0.5407,0.7577,8.0200
... ...

2.YOLOv5

(venv_fasterrcnn) sunjinge@ares1:~/PyTorchModels/Detection/yolo$ python yolo_log_parser.py -l yolov5_train.log -o yolov5_metrics.csv
✅ 解析完成！结果已保存到 yolov5_metrics.csv
共处理 19 个epoch
(venv_fasterrcnn) sunjinge@ares1:~/PyTorchModels/Detection/yolo$ cat yolov5_metrics.csv
epoch,gpu_mem,box_loss,obj_loss,cls_loss,instances
0,6.419968,0.100159,0.083691,0.096369,212.332793
1,6.420000,0.073442,0.084803,0.073568,211.711195
... ...

3.SSD

(venv_fasterrcnn) sunjinge@ares1:~/PyTorchModels/Detection/ssd$ python ssd_log_parser.py -l ssd_train.log -o ssd_metrics.csv
解析完成，结果已保存到 ssd_metrics.csv
(venv_fasterrcnn) sunjinge@ares1:~/PyTorchModels/Detection/ssd$ cat ssd_metrics.csv
epoch,batch,batch_time,data_time,loss
0,0/2069,17.007,0.434,24.2514
1,0/2069,3.336,0.4,5.3279
... ...