Closed liqima closed 5 years ago
You just not need to build from source, you need to build from source with ffmpeg and opencv. Maybe, you can ask for a precompiled caffe2 from someone who has compiled it from source with ffmpeg and opencv.
Build from source with USE_FFMPEG and USE_OPENCV is ON should solve this.
Hi, I installed caffe2 with anaconda, and something wrong when I run tools/extract_features.py
Traceback (most recent call last): File "extract_features.py", line 334, in <module> main() File "extract_features.py", line 329, in main ExtractFeatures(args) File "extract_features.py", line 134, in ExtractFeatures devices=gpus, File "/home/maliqi/anaconda2/envs/detect/lib/python2.7/site-packages/caffe2/python/data_parallel_model.py", line 34, in Parallelize_GPU Parallelize(*args, **kwargs) File "/home/maliqi/anaconda2/envs/detect/lib/python2.7/site-packages/caffe2/python/data_parallel_model.py", line 231, in Parallelize input_builder_fun(model_helper_obj) File "extract_features.py", line 109, in input_fn use_local_file=args.use_local_file, File "/home/maliqi/video_repetition/utils/model_helper.py", line 127, in AddVideoInput data, label = model.net.VideoInput( File "/home/maliqi/anaconda2/envs/detect/lib/python2.7/site-packages/caffe2/python/core.py", line 2179, in __getattr__ ",".join(workspace.C.nearby_opnames(op_type)) + ']' AttributeError: Method VideoInput is not a registered operator. Did you mean: []
I find some discuss here and here, sound like I have to use caffe2 installed from source. But unfortunatly, my machine cannot do that (I know how to compile install caffe2, here I just can't do that), so, I want to ask is there another way to solve this problem? thanks.