djdam / faster-rcnn-scenarios Goto Github PK

I ran the ananlysis script and got a weird result as below, any idea about what goes wrong ?

min im width: 6.00
mean im width: 33.49
max im width: 188.00
min im height: 6.00
mean im height: 33.49
max im height: 188.00
min scale_factors: 3.19
mean scale_factors: 34.98
max scale_factors: 100.00
min bbox width: 38.00
mean bbox width: 982.76
max bbox width: 12400.00
min bbox height: 39.00
mean bbox height: 982.76
max bbox height: 12400.00
min aspect ratio: 1.00
mean aspect ratio: 1.00
max aspect ratio: 1.00

Hello!

For a project I want to detect small images with faster rcnn. I found out that I have to change the variables in the function of the first_stage_anchor_generator in the config file. Via a forum I found your link to imdb_analyse.py, which can help me find these values.

To run this, I set my environment variable
FASTER_RCNN_ROOT: C:\fasterrcnn\fast-rcnn-master\fast-rcnn-master

This leads to the folder including the following folders: caffe-fast-rcnn, data, experiments, lib, matlab, models, output, tools.

Is this the right folder I have to redirect my FASTER_RCNN_ROOT variable to? And next to that, the requirements say that CAFFE_ROOT needs to be set too, but I do not know where to find this folder.

When I run the program now, I get the following error: ModuleNotFoundError: No module named '_init_paths'. I do not know how to run your program exactly. Does the .py need to be inside the folder with images(jpg) and xml files? Can you help me out? It would really help me.

(I set the python path and this worked before)

Thanks in advance,

Niek

请问如何将环境变量CAFFE_ROOT设置为指向本地caffe安装？