lxrd-aj / yolo_v1

You Only Look Once version 1

YOLOv1

A technical article and PyTorch Implementation of the popular YOLO object detection algorithm by Joseph Redmon. Follow this link for the technical article available online at AraIntelligence.com which provides an in-depth explanation about the algorithm.

PIP Requirements

The python version used for development is Python 3.6.10 and the libraries used are located in requirements.txt. Highlighted below are the most important ones

• numpy==1.18.1
• Pillow==7.0.0
• torch==1.5.0
• torchvision==0.6.0
• torchviz==0.0.1

Training

The model was trained on the Pascal VOC 2007+2012 dataset which was partitioned into a training set of 13,170 images and a validation set of 8,333 images. There are some helper scripts available:

• ./data/download_voc.sh : Downloads the VOC dataset from Joseph Redmon's VOC mirror and partitions the dataset into a training and validation one
• python training.py : is the script used to train the model

The model achieves a relatively low training loss on the dataset. The validation loss is lower than the training loss due to the data augmentation applied to the train dataset during training.

Image Augmentation

Sample augmentations applied are shown below

Color Jitter

Random Blur

Random Horizontal Flip

Random Vertical Flip

Inference

Inference speed

Device	Backbone	Input Size	Time (seconds)
2.9 GHz Dual-Core Intel Core i5	ResNet50	448x448	2s
NVIDIA Tesla K80 GPU	ResNet50	448x448	-

Prediction

• Image File on disk: python detect_image --image="./data/car_bike.png" --model="./model_checkpoints/90_epoch.pth"
• Video file on disk: python detect_image --video="./data/IMG_4855.mov" --model="./model_checkpoints/90_epoch.pth" --output="./processed_IMG_4855.mov"

yolo_v1's Issues

why in def forward(self, x) function have "exit(0)"

hello, Lxrd-AJ:
I have detail read you blog in "YOLOv1: You Only Look Once"(https://araintelligence.com/blogs/deep-learning/object-detection/yolo_v1/)
but when i read the yolov1.py, i can't understant why have "exit(0)" in the function of "def forward(self, x)", because i understand that the function "exit(0)" is exit the program, could give me a more detail explain?

thankyou a lot!

crash

hi Lxrd-AJ:
I run your code, but when one epoch will end, its output error as below
my torch version is: 1.6.0

Traceback (most recent call last):
File "xxxx/python-code/yolov1/training.py", line 157, in
predictions = model(images)
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/data_parallel.py", line 155, in forward
outputs = self.parallel_apply(replicas, inputs, kwargs)
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/data_parallel.py", line 165, in parallel_apply
return parallel_apply(replicas, inputs, kwargs, self.device_ids[:len(replicas)])
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/parallel_apply.py", line 85, in parallel_apply
output.reraise()
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/_utils.py", line 395, in reraise
raise self.exc_type(msg)
ValueError: Caught ValueError in replica 0 on device 0.
Original Traceback (most recent call last):
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/parallel_apply.py", line 60, in _worker
output = module(*input, **kwargs)
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "xxxx/python-code/yolov1/yolo_v1.py", line 70, in forward
linearOutput = self.linear_layers(flattened)
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/modules/container.py", line 117, in forward
input = module(input)
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/modules/batchnorm.py", line 131, in forward
return F.batch_norm(
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/functional.py", line 2012, in batch_norm
_verify_batch_size(input.size())
File "xxxx/anaconda3/lib/python3.8/site-packages/torch/nn/functional.py", line 1995, in _verify_batch_size
raise ValueError('Expected more than 1 value per channel when training, got input size {}'.format(size))
ValueError: Expected more than 1 value per channel when training, got input size torch.Size([1, 12544])