jwyang / faster-rcnn.pytorch

A faster pytorch implementation of faster r-cnn
MIT License
7.63k stars 2.34k forks source link

python3.7,cuda10,pytorch1.1 #620

Open MaLiXiaoGuo opened 4 years ago

MaLiXiaoGuo commented 4 years ago

Traceback (most recent call last): File "F:/faster-rcnn.pytorch-master/demo.py", line 27, in from roi_data_layer.roidb import combined_roidb File "F:\faster-rcnn.pytorch-master\lib\roi_data_layer\roidb.py", line 12, in from datasets.factory import get_imdb File "F:\faster-rcnn.pytorch-master\lib\datasets\factory.py", line 14, in from datasets.pascal_voc import pascal_voc File "F:\faster-rcnn.pytorch-master\lib\datasets\pascal_voc.py", line 23, in from .imdb import imdb File "F:\faster-rcnn.pytorch-master\lib\datasets\imdb.py", line 14, in from model.utils.cython_bbox import bbox_overlaps ModuleNotFoundError: No module named 'model.utils.cython_bbox'

EMCP commented 4 years ago

this is patched by replacing the pycocotools with the latest from github

example at

https://github.com/jwyang/faster-rcnn.pytorch/pull/466

EMCP commented 4 years ago

my tutorial to get this working is at https://mahr.io/tutorials/faster-r-cnn-pytorch-how-to

harishkashyap commented 2 years ago

The tutorial isn't working anymore. How I so wish, people building their modules on pytorch also update their libraries to work with latest versions in general.

EMCP commented 2 years ago

@harishkashyap this entire project which the tutorial was based on became out of date. Little reason to attempt to update what requires proper refactoring by the original author who clearly hasn't touched this repo in a while. I've personally moved into torchvision and pytorch 1.9 in order to attempt to do object detection on custom data

https://pytorch.org/tutorials/intermediate/torchvision_tutorial.html

My full time job has not allowed me time to work beyond the basics

harishkashyap commented 2 years ago

Appreciate yoir quick response on this. I certainly understand given the way pytorch keeps updating with breaking changes. Tensorflow is worse. This repo would be incredibly useful if it works out of the box. Great work indeed!