未验证 提交 40ff9a63 编写于 作者: Y Yang Zhang 提交者: GitHub

Add inference support for efficientdet (#1345)

reuse RetinaNet pipeline for now
上级 929c7f20
......@@ -78,6 +78,15 @@ def parse_reader(reader_cfg, metric, arch):
params['image_shape'] = image_shape[1:]
if 'target_dim' in params:
params.pop('target_dim')
if p['type'] == 'ResizeAndPad':
assert has_shape_def, "missing input shape"
p['type'] = 'Resize'
p['target_size'] = params['target_dim']
p['max_size'] = params['target_dim']
p['interp'] = params['interp']
p['image_shape'] = image_shape[1:]
preprocess_list.append(p)
continue
p.update(params)
preprocess_list.append(p)
batch_transforms = reader_cfg.get('batch_transforms', None)
......@@ -113,9 +122,9 @@ def dump_infer_config(FLAGS, config):
'Face': 3,
'TTFNet': 3,
'FCOS': 3,
'EfficientDet': 40
}
infer_arch = config['architecture']
infer_arch = 'RetinaNet' if infer_arch == 'EfficientDet' else infer_arch
for arch, min_subgraph_size in trt_min_subgraph.items():
if arch in infer_arch:
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册