运行检测模型卷积通道剪裁示例时报错

[liu36259069]

当我运行python compress.py
-s yolov3_mobilenet_v1_slim.yaml
-c ../../configs/yolov3_mobilenet_v1.yml
-o max_iters=20
num_classes=4
YoloTrainFeed.batch_size=32
pretrain_weights=/home/aistudio/PaddleDetection/output/yolov3_mobilenet_v1/best_model
-d "/home/aistudio/work/coco"
这个命令时一直报错以下错误，经定位是eval_utils.py将gt_box的值转换成im_id异常，异常信息如下

loading annotations into memory...
Done (t=0.01s)
creating index...
index created!
[ 16. 115. 218. 374.]
Traceback (most recent call last):
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/contrib/slim/core/compressor.py", line 593, in run
self._eval(context)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/contrib/slim/core/compressor.py", line 542, in _eval
func(self.eval_graph.program, self.scope))
File "compress.py", line 207, in eval_func
FLAGS.output_eval)
File "../../ppdet/utils/eval_utils.py", line 205, in eval_results
is_bbox_normalized=is_bbox_normalized)
File "../../ppdet/utils/coco_eval.py", line 86, in bbox_eval
results, clsid2catid, is_bbox_normalized=is_bbox_normalized)
File "../../ppdet/utils/coco_eval.py", line 215, in bbox2out
im_id = int(im_ids[i][0])
TypeError: only size-1 arrays can be converted to Python scalars
2019-11-20 20:32:00,491-ERROR: None
2019-11-20 20:32:00,491-ERROR: None
2019-11-20 20:32:01,633-INFO: epoch:1; batch_id:0; odict_keys(['loss', 'lr']) = [117.678, 0.0]

后面通过排查发现compress.py 将gt_box的值当成im_id出入导致报错，
在第79行新增outs.append(data['im_id']) 解决问题

[qingqing01]

@liu36259069 感谢反馈，我们修复下~