NOTE: Due to errors in Credit/Debit cards or errors in AWS/Google cloud platforms i was not able to deploy the application. So, I just created API & Dockerized the Application. Further Implementation of project is below. Please go through it.
- Table of Contents
- About
- YOLOV7
- 📁 Datasets
- 📖 Data Preprocessing
- Requirements
- Installation
- 🚀 Setup (On Production Machine)
- 💡 How to Run
- Additional guides
- Documentation
- 📂 Directory Tree
- 🔑 Results
- 👏 And it's done!
- ❤️ Owner
- 👀 License
Monkey Detection using Computer Vision using AI. It uses live video stream from camera feed or rtsp streaming from ip camera or cctv or input images and use object detection to detect Monkey in these feeds. This project is part of Recruitment as an Internship in IIT Bombay. In this project I used YOLOV7 model to detect Monkey in the video feed.
MS COCO
| Model | Test Size | APtest | AP50test | AP75test | batch 1 fps | batch 32 average time |
|---|---|---|---|---|---|---|
| YOLOv7 | 640 | 51.4% | 69.7% | 55.9% | 161 fps | 2.8 ms |
| YOLOv7-X | 640 | 53.1% | 71.2% | 57.8% | 114 fps | 4.3 ms |
| YOLOv7-W6 | 1280 | 54.9% | 72.6% | 60.1% | 84 fps | 7.6 ms |
| YOLOv7-E6 | 1280 | 56.0% | 73.5% | 61.2% | 56 fps | 12.3 ms |
| YOLOv7-D6 | 1280 | 56.6% | 74.0% | 61.8% | 44 fps | 15.0 ms |
| YOLOv7-E6E | 1280 | 56.8% | 74.4% | 62.1% | 36 fps | 18.7 ms |
Docker environment (recommended)
Expand
# create the docker container, you can change the share memory size if you have more.
nvidia-docker run --name yolov7 -it -v your_coco_path/:/coco/ -v your_code_path/:/yolov7 --shm-size=64g nvcr.io/nvidia/pytorch:21.08-py3
# apt install required packages
apt update
apt install -y zip htop screen libgl1-mesa-glx
# pip install required packages
pip install seaborn thop
# go to code folder
cd /yolov7yolov7.pt yolov7x.pt yolov7-w6.pt yolov7-e6.pt yolov7-d6.pt yolov7-e6e.pt
python test.py --data data/coco.yaml --img 640 --batch 32 --conf 0.001 --iou 0.65 --device 0 --weights yolov7.pt --name yolov7_640_valYou will get the results:
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.51206
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.69730
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.55521
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.35247
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.55937
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.66693
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.38453
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.63765
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.68772
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.53766
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.73549
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.83868
To measure accuracy, download COCO-annotations for Pycocotools to the ./coco/annotations/instances_val2017.json
Data preparation
bash scripts/get_coco.sh- Download MS COCO dataset images (train, val, test) and labels. If you have previously used a different version of YOLO, we strongly recommend that you delete
train2017.cacheandval2017.cachefiles, and redownload labels
Single GPU training
# train p5 models
python train.py --workers 8 --device 0 --batch-size 32 --data data/coco.yaml --img 640 640 --cfg cfg/training/yolov7.yaml --weights '' --name yolov7 --hyp data/hyp.scratch.p5.yaml
# train p6 models
python train_aux.py --workers 8 --device 0 --batch-size 16 --data data/coco.yaml --img 1280 1280 --cfg cfg/training/yolov7-w6.yaml --weights '' --name yolov7-w6 --hyp data/hyp.scratch.p6.yamlMultiple GPU training
# train p5 models
python -m torch.distributed.launch --nproc_per_node 4 --master_port 9527 train.py --workers 8 --device 0,1,2,3 --sync-bn --batch-size 128 --data data/coco.yaml --img 640 640 --cfg cfg/training/yolov7.yaml --weights '' --name yolov7 --hyp data/hyp.scratch.p5.yaml
# train p6 models
python -m torch.distributed.launch --nproc_per_node 8 --master_port 9527 train_aux.py --workers 8 --device 0,1,2,3,4,5,6,7 --sync-bn --batch-size 128 --data data/coco.yaml --img 1280 1280 --cfg cfg/training/yolov7-w6.yaml --weights '' --name yolov7-w6 --hyp data/hyp.scratch.p6.yamlyolov7_training.pt yolov7x_training.pt yolov7-w6_training.pt yolov7-e6_training.pt yolov7-d6_training.pt yolov7-e6e_training.pt
Single GPU finetuning for custom dataset
# finetune p5 models
python train.py --workers 8 --device 0 --batch-size 32 --data data/custom.yaml --img 640 640 --cfg cfg/training/yolov7-custom.yaml --weights 'yolov7_training.pt' --name yolov7-custom --hyp data/hyp.scratch.custom.yaml
# finetune p6 models
python train_aux.py --workers 8 --device 0 --batch-size 16 --data data/custom.yaml --img 1280 1280 --cfg cfg/training/yolov7-w6-custom.yaml --weights 'yolov7-w6_training.pt' --name yolov7-w6-custom --hyp data/hyp.scratch.custom.yamlOn video:
python detect.py --weights yolov7.pt --conf 0.25 --img-size 640 --source yourvideo.mp4On image:
python detect.py --weights yolov7.pt --conf 0.25 --img-size 640 --source inference/images/horses.jpgThis dataset is downloaded from roboflow using YOLOV7 model. You can find the data Here.
Data pre-processing is an important step for the creation of a machine learning model. Initially, data may not be clean or in the required format for the model which can cause misleading outcomes. In pre-processing of data, we transform data into our required format. It is used to deal with noises, duplicates, and missing values of the dataset. Data pre-processing has the activities like importing datasets, splitting datasets, attribute scaling, etc. Preprocessing of data is required for improving the accuracy of the model.
System requirements depends on complexity or size of object detection model, larger model will require more compute power and will be good at detection. I have used this in Google Colab with Tesla T4 GPU & Kaggle with Tesla p100 GPU which is a good device. With my Nvidia Geforce 330 MX GPU based system it is not sufficient to train a YOLOV7 model.
Minimum of 4GB NVIDIA Graphics Card is required to train YOLOV7 model.
Python 3.6 or higher. Tested with Python 3.6, 3.7, 3.8, 3.9, 3.10 in Windows 11 and Linux.
Torch>= 1.7.0numpy>= 1.18.5opencv-python>= 4.1.1
This implementation is tested with PyTorch gpu in Windows 11 and Linux and Google Colab
All the dependencies and required libraries are included in the file requirements.txt See here
There are two ways to install python in windows using Python 3 installer or Anaconda. Installing python with anaconda or miniconda is recommended. In linux Python 3 is installed by default but we can also install miniconda or conda into linux.
Create a new python virtual environment using conda or venv and activate it. If Anaconda or Miniconda is installed use conda else use venv to create virtual environments.
- Using conda
conda create --name security_cam
conda activate security_cam
conda install pip- Using venv in linux
python3 -m venv path/to/create/env/security_cam
source path/to/create/env/security_cam/bin/activate- Using venv in windows
python -m venv path\to\create\env\security_cam
path\to\create\env\security_cam\Scripts\activateThe command below will install all the required dependencies from requirements.txt file.
pip install -r requirements.txtDownload the YOLOV7 model from here and place it in the root directory of the project. Then copy the dataset you downloaded from Roboflow to the YOLOV7-MAIN Folder and add the data.yaml file in main directory of the project. The data.yaml file should contain the following information: (The dataset folder name is monkey_data)
Note: Clone the YOLOV7 repository and do not download as a zip since .git folder is required for training and create a "static" and "results" named empty folder after copying templates folder to YOLOV7.
git clone https://github.com/WongKinYiu/yolov7.gitThe data.yaml file should contain the following information:
train: monkey_data/train/images
val: monkey_data/valid/images
nc: 1
names: ["monkey"]- Change your directory to the cloned repo
cd "yolov7-main"- Also copy the files from this repository to the YOLO Folder same as the structured folder in the given drive link.
-
I used "Flask" as an API to connect the model with the web application.
-
Check out the flask file here app.py
-
Then I created a Docker image using Dockerfile and pushed it to the Docker Hub.
-
We can also use Docker-compose to run the application.
-
Due to errors in Credit/Debit cards or errors in AWS/Google cloud platforms i was not able to deploy the application.
- Open terminal. Go into the cloned project directory and type the following command:
python3 train.py --epochs 128 --workers 8 --device 0 --batch-size 32 --data data.yaml --img 416 416 --cfg cfg/training/yolov7.yaml --weights '' --name Monkey-Model --hyp data/hyp.scratch.p5.yaml- After training is done, run the following command to test the model on a video/image:
py detect.py --weights best.pt --project results --name results --source static/img.jpg --save-txt --save-conf- If you want to try the Project then run the flask file app.py
python app.pyIf get stuck in installation part follow these additional resources
The Input Parameters used in this project are
- --epochs: Number of epochs to train the model.
- --workers: Number of workers to use for data loading.
- --device: Device to use for training.(0 for GPU and 1 for CPU)
- --batch-size: Batch size for training.
- --data: Path to data.yaml file.
- img: Image size.
- --cfg: Path to model.yaml file.
- --weights: Path to weights file. (Leave blank for training from scratch)
- --name: Name of the model.
- --hyp: Path to hyperparameters file.
Mostly you are supposed to get errors regarding cv2.imwrite() to a particular folder, At that time you need to create respected folders in the main directory.
├───cfg
│ ├───baseline
│ ├───deploy
│ └───training
├───data
├───deploy
│ └───triton-inference-server
│ └───data
├───models
│ └───__pycache__
├───monkey_data
│ ├───test
│ │ ├───images
│ │ └───labels
│ ├───train
│ │ ├───images
│ │ └───labels
│ └───valid
│ ├───images
│ └───labels
├───results
├───runs
│ └───train
│ └───Monkey-Model
│ └───weights
├───scripts
├───static
├───templates
├───tools
└───utils
├───aws
├───google_app_engine
├───wandb_logging
└───__pycache__
Our YOLOV7 Model gives the best accuracy and some good F1 Scores etc., & you can see below :
F1 Score:
P Curve:
PR Curve:
R Curve:
Results:
Test Case 1:
Test Case 2:
Test Case 3:
Feel free to mail me for any doubts/query 📧 chagantivenkataramireddy1@gmail.com
You are allowed to cite any part of the code or our dataset. You can use it in your Research Work or Project. Remember to provide credit to the Maintainer Chaganti Reddy by mentioning a link to this repository and her GitHub Profile.
Follow this format:
- Author's name - Chaganti Reddy
- Date of publication or update in parentheses.
- Title or description of document.
- URL.
You can find the references I used in the references.md file.
Made with ❤️ by Chaganti Reddy
MIT © Chaganti Reddy