Skip to content

Latest commit

 

History

History
144 lines (123 loc) · 4.6 KB

README.md

File metadata and controls

144 lines (123 loc) · 4.6 KB
Raw

NeuroFind

A solution to the task given by the Oberseminar of Messtechnik Institute of TU Dresden in 2021

alt text

Installation

please install on a linux operating system

Install both Faster R-CNN and Yolov5

git clone https://github.com/Anxum/NeuroFind.git
cd NeuroFind
git clone https://github.com/ultralytics/yolov5.git temp
mv temp/.git yolov5/.git
mv temp/* yolov5
rm -rf temp
cd yolov5
pip3 install -r requirements.txt
cd ../Faster_R-CNN
git clone https://github.com/tensorflow/models.git
pip3 install lxml
mkdir models/protoc
mv protoc-3.19.3-linux-x86_64.zip models/protoc
cd models/protoc
unzip protoc-3.19.3-linux-x86_64.zip
rm -r protoc-3.19.3-linux-x86_64.zip
cd ../research
../protoc/bin/protoc object_detection/protos/*.proto --python_out=.
export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim
cp object_detection/packages/tf2/setup.py .
python3 -m pip install --use-feature=2020-resolver .
cd ../../../..

Only Install Yolov5

git clone https://github.com/Anxum/NeuroFind.git
cd NeuroFind
git clone https://github.com/ultralytics/yolov5.git temp
mv temp/.git yolov5/.git
mv temp/* yolov5
rm -rf temp
cd yolov5
pip3 install -r requirements.txt 
cd ../..

Only Install Faster R-CNN

git clone https://github.com/Anxum/NeuroFind.git
cd NeuroFind/Faster_R-CNN
git clone https://github.com/tensorflow/models.git
pip3 install lxml
mkdir models/protoc
mv protoc-3.19.3-linux-x86_64.zip models/protoc
cd models/protoc
unzip protoc-3.19.3-linux-x86_64.zip
rm -r protoc-3.19.3-linux-x86_64.zip
cd ../research
../protoc/bin/protoc object_detection/protos/*.proto --python_out=.
export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim
cp object_detection/packages/tf2/setup.py .
python3 -m pip install --use-feature=2020-resolver .
cd ../../../..

Weights

In addition to the installation guide above please download the weights for the neuronal networks from: --insert url--

Standardized paths for YOLOv5 and Faster R-CNN:

Neuronal Network Put Weights to:
YOLOv5 NeuroFind/yolov5/runs/train/weights
Faster R-CNN NeuroFind/Faster_R-CNN/export

If you choose to put the weights into another folder you must refer to these paths if you wish to use the inference algorithms later.

Inference

Please download weigts first as described above or train the network yourself before Inference

YOLOv5

For Inference please run (from NeuroFind/yolov5)

python3 yolo_detect.py

Possible arguments include:

Argument Description Input format Default Value Comment
--weights Put path to weights OS_PATH ./runs/train/weights/best.pt
--source path to file or dir OS_PATH ../Images URL, glob and webcam(0) could be possible, not tested
--conf-thres minimal value of confidence for bounding boxes float, [0..1] 0.25
--device Specify the device to run Inference 0,1,2,3,..., cpu for CPU
--nosave Don't save the images after inference False
--project Save to project/name OS_PATH ./runs/detect
--name Save to project/name str exp
--line-thickness Thickness of lines of bounding boxes on the image int 2

Arguments that are not tested or will not work or should not be specified

Not testet Will not work Please don't use
--exist-ok --visualize --hide labels
--classes --view-img --hide-conf
--half --save-conf
--dnn --imgsz
--update --save-txt
--augment
--agnostic-nms
--iou-thres
--max-det

Faster R-CNN

For Inference please run (from NeuroFind/Faster_R-CNN)

python3 frcnn_detect.py

Possible arguments include:

Argument Description Input format Default Value
--config Put path to config OS_PATH ./export/pipeline.config
--source path to file or dir OS_PATH ../Images
--checkpt Path to checkpoint OS_PATH ./export/checkpoint/ckpt-0
--labelmap Path to labelmap OS_PATH ./export/label_map.pbtxt
--conf-thres minimal value of confidence for bounding boxes float, [0..1] 0.25
--output Save files to this path OS_PATH ./runs/detect/exp
--nosave Don't save the images after inference False
--line-thickness Thickness of lines of bounding boxes on the image int 2

Training

YOLOv5

to create a training dataset please use( in NeuroFind/training_data_yolo):

python3 crop_image.py

For training, please refer to: https://github.com/ultralytics/yolov5.git

Faster R-CNN

For training, please refer to: https://github.com/tensorflow/models.git