darknet - YOLOv3 训练过程中报错 (Cannot load image "COCO_train2014_000000167126.jpg"; STB Reason: expected marker; darknet: ./src/cuda.c:36: check_error: Assertion `0' failed.)
1. 训练日志
...... Resizing 352 Loaded: 0.863524 seconds Region 82 Avg IOU: 0.813832, Class: 0.464163, Obj: 0.205051, No Obj: 0.005099, .5R: 1.000000, .75R: 0.750000, count: 4 Region 82 Avg IOU: 0.498435, Class: 0.830196, Obj: 0.018140, No Obj: 0.001208, .5R: 0.000000, .75R: 0.000000, count: 1 Region 82 Avg IOU: 0.674023, Class: 0.662911, Obj: 0.043203, No Obj: 0.004061, .5R: 1.000000, .75R: 0.333333, count: 3 Region 82 Avg IOU: 0.728529, Class: 0.804345, Obj: 0.365087, No Obj: 0.003158, .5R: 1.000000, .75R: 0.000000, count: 1 Region 82 Avg IOU: 0.605464, Class: 0.518218, Obj: 0.031066, No Obj: 0.002984, .5R: 1.000000, .75R: 0.000000, count: 2 Region 82 Avg IOU: 0.672544, Class: 0.460274, Obj: 0.214596, No Obj: 0.009742, .5R: 0.777778, .75R: 0.444444, count: 9 Region 82 Avg IOU: 0.775507, Class: 0.668895, Obj: 0.123877, No Obj: 0.004137, .5R: 1.000000, .75R: 0.750000, count: 4 Region 82 Avg IOU: 0.717642, Class: 0.015938, Obj: 0.027864, No Obj: 0.004484, .5R: 1.000000, .75R: 0.400000, count: 5 Region 82 Avg IOU: 0.793616, Class: 0.368892, Obj: 0.335050, No Obj: 0.008201, .5R: 1.000000, .75R: 1.000000, count: 3 Region 82 Avg IOU: 0.873850, Class: 0.974957, Obj: 0.901814, No Obj: 0.006406, .5R: 1.000000, .75R: 1.000000, count: 2 Region 82 Avg IOU: 0.748278, Class: 0.925417, Obj: 0.209667, No Obj: 0.004293, .5R: 1.000000, .75R: 0.750000, count: 4 Region 82 Avg IOU: 0.523779, Class: 0.975091, Obj: 0.081971, No Obj: 0.000979, .5R: 1.000000, .75R: 0.000000, count: 1 Cannot load image "/home/strong/darknet_work/darknet_benchmark/darknet/coco_dataset/coco/images/train2014/COCO_train2014_000000167126.jpg" STB Reason: expected marker CUDA Error: driver shutting down darknet: ./src/cuda.c:36: check_error: Assertion `0' failed. strong@foreverstrong:~/darknet_work/darknet_180508/darknet$
2. 问题分析和解决
打开如下路径的数据，发现训练图片存在问题。
/home/strong/darknet_work/darknet_benchmark/darknet/coco_dataset/coco/images/train2014/COCO_train2014_000000167126.jpg
1. 删除问题图片，并在训练图片列表 (trainvalno5k.txt) 中删除该图片路径。
或者
1. 查看是否数据集下载问题，重新下载数据集替换问题图片。