YOLOv4 Object Detection API Server
This repository is mixture version of https://github.com/theAIGuysCode/Object-Detection-API (YOLOv3 REST API) and https://github.com/theAIGuysCode/tensorflow-yolov4-tflite (YOLOv4). so that I added REST API functions on YOLOv4.
YOLOv4, YOLOv4-tiny Implemented in Tensorflow 2.0. Convert YOLO v4, YOLOv3, YOLO tiny .weights to .pb, .tflite and trt format for tensorflow, tensorflow lite, tensorRT, Flask, REST API.
1. Install
Conda (Recommended)
# Tensorflow CPU
conda env create -f conda-cpu.yml
conda activate yolov4-cpu
# Tensorflow GPU
conda env create -f conda-gpu.yml
conda activate yolov4-gpu
# For apple silicon mac
conda env create -f conda-gpu-apple-silicon-mac.yml
conda activate yolov4-gpuPip
# TensorFlow CPU
pip install -r requirements.txt
# TensorFlow GPU
pip install -r requirements-gpu.txtDownloading Official Pre-trained Weights
YOLOv4 comes pre-trained and able to detect 80 classes. For easy demo purposes we will use the pre-trained weights. Download pre-trained yolov4.weights file: https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT
Copy and paste yolov4.weights from your downloads folder into the 'data' folder of this repository.
If you want to use yolov4-tiny.weights, a smaller model that is faster at running detections but less accurate, download file here: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.weights
Using Custom Trained YOLOv4 Weights
Learn How To Train Custom YOLOv4 Weights here: https://www.youtube.com/watch?v=mmj3nxGT2YQ
Copy and paste your custom .weights file into the 'data' folder and copy and paste your custom .names into the 'data/classes/' folder.
The only change within the code you need to make in order for your custom model to work is on line 14 of 'core/config.py' file.
Update the code to point at your custom .names file as seen below. (my custom .names file is called custom.names but yours might be named differently)
Note: If you are using the pre-trained yolov4 then make sure that line 14 remains coco.names.
Convert
To implement YOLOv4 using TensorFlow, first we convert the .weights into the corresponding TensorFlow model files and then run the model.
# Convert darknet weights to tensorflow
## yolov4
python save_model.py --weights ./data/yolov4.weights --output ./checkpoints/yolov4-416 --input_size 416 --model yolov4
# yolov4-tiny
python save_model.py --weights ./data/yolov4-tiny.weights --output ./checkpoints/yolov4-tiny-416 --input_size 416 --model yolov4 --tiny
# custom yolov4
python save_model.py --weights ./data/custom.weights --output ./checkpoints/custom-416 --input_size 416 --model yolov4 Run
Run Flask app
python app.pyRun Waitress wsgi for production deployment
python waitress-server.py2. Usage
Detections by image files (POST http://localhost:5050/detections/by-image-files)
need more than 0 image files, multipart/form-data, key name is "images"
Request example(python.requests)
import requests
url = "http://127.0.0.1:5050/detections/by-image-files"
payload={}
files=[
('images',('dog.jpg',open('/C:/Users/Qone/repos/YOLOv4-Object-Detection-API-Server/data/images/dog.jpg','rb'),'image/jpeg')),
('images',('kite.jpg',open('/C:/Users/Qone/repos/YOLOv4-Object-Detection-API-Server/data/images/kite.jpg','rb'),'image/jpeg'))
]
headers = {}
response = requests.request("POST", url, headers=headers, data=payload, files=files)
print(response.text)Response example
{
"response": [
{
"image": "dog.jpg",
"detections": [
{
"class": "bicycle",
"confidence": 98.68,
"box": [
0.22659872472286224,
0.16641898453235626,
0.7383183240890503,
0.7395057082176208
]
},
{
"class": "dog",
"confidence": 98.52,
"box": [
0.4033146798610687,
0.1720491647720337,
0.9337449669837952,
0.406764417886734
]
},
{
"class": "truck",
"confidence": 92.15,
"box": [
0.1300187110900879,
0.6068406701087952,
0.2994907796382904,
0.9028187990188599
]
}
]
},
{
"image": "kite.jpg",
"detections": [
{
"class": "person",
"confidence": 99.11,
"box": [
0.7725002765655518,
0.15819524228572845,
0.955871045589447,
0.20016619563102722
]
},
{
"class": "kite",
"confidence": 98.88,
"box": [
0.08883842825889587,
0.4350317120552063,
0.17153428494930267,
0.4975464642047882
]
},
{
"class": "person",
"confidence": 98.6,
"box": [
0.6820783615112305,
0.08408961445093155,
0.84914231300354,
0.1225830614566803
]
},
{
"class": "kite",
"confidence": 87.96,
"box": [
0.38221558928489685,
0.42745858430862427,
0.40842822194099426,
0.4442833960056305
]
},
{
"class": "person",
"confidence": 85.87,
"box": [
0.5421252250671387,
0.2548080086708069,
0.560500979423523,
0.26386842131614685
]
},
{
"class": "kite",
"confidence": 84.4,
"box": [
0.2625654339790344,
0.20689727365970612,
0.3109365999698639,
0.22676025331020355
]
},
{
"class": "person",
"confidence": 78.57,
"box": [
0.5600144267082214,
0.38356536626815796,
0.585326611995697,
0.39563310146331787
]
},
{
"class": "person",
"confidence": 78.54,
"box": [
0.5680103302001953,
0.02381865493953228,
0.6170229911804199,
0.042786046862602234
]
},
{
"class": "person",
"confidence": 76.38,
"box": [
0.5647990107536316,
0.059801846742630005,
0.6308582425117493,
0.07934856414794922
]
},
{
"class": "kite",
"confidence": 73.57,
"box": [
0.37724360823631287,
0.3463357090950012,
0.3972048759460449,
0.3593466877937317
]
},
{
"class": "person",
"confidence": 68.83,
"box": [
0.5980629920959473,
0.13111814856529236,
0.6402380466461182,
0.14322905242443085
]
},
{
"class": "kite",
"confidence": 68.4,
"box": [
0.43732577562332153,
0.8011485934257507,
0.47132566571235657,
0.8150034546852112
]
},
{
"class": "kite",
"confidence": 55.76,
"box": [
0.41698184609413147,
0.22577570378780365,
0.4412486255168915,
0.24203325808048248
]
},
{
"class": "person",
"confidence": 46.35,
"box": [
0.5745782852172852,
0.3943392336368561,
0.5909861326217651,
0.4089730978012085
]
},
{
"class": "person",
"confidence": 41.32,
"box": [
0.5020405650138855,
0.891931414604187,
0.5144351124763489,
0.899072527885437
]
},
{
"class": "kite",
"confidence": 40.03,
"box": [
0.4195924699306488,
0.5634137988090515,
0.4349488317966461,
0.5703240036964417
]
},
{
"class": "person",
"confidence": 36.76,
"box": [
0.5806989669799805,
0.019801238551735878,
0.6189168095588684,
0.037031080573797226
]
}
]
}
]
}Image by image file(POST http://localhost:5050/image/by-image-file)
need 1 image file, multipart/form-data, key name is "images"
Request example(nodejs.axios)
var axios = require('axios');
var FormData = require('form-data');
var fs = require('fs');
var data = new FormData();
data.append('images', fs.createReadStream('/C:/Users/Qone/repos/YOLOv4-Object-Detection-Server/data/images/dog.jpg'));
var config = {
method: 'post',
url: 'http://127.0.0.1:5050/image/by-image-file',
headers: {
...data.getHeaders()
},
data : data
};
axios(config)
.then(function (response) {
console.log(JSON.stringify(response.data));
})
.catch(function (error) {
console.log(error);
});Response example
Detections by image url list(POST http://localhost:5050/detections/by-url-list)
need image url list json form
Request example
{
"images" : [
"https://images.unsplash.com/photo-1627662055794-94ab33f5913a",
"https://images.unsplash.com/photo-1622495506073-56b1152a010c"
]
}Response example
{
"response": [
{
"image": "Image1",
"detections": [
{
"class": "bowl",
"confidence": 98.49,
"box": [
0.8068702220916748,
0.4503408670425415,
0.9296261072158813,
0.5750141143798828
]
},
{
"class": "cup",
"confidence": 97.94,
"box": [
0.5713980197906494,
0.6671542525291443,
0.7300602197647095,
0.7218816876411438
]
},
{
"class": "cup",
"confidence": 97.69,
"box": [
0.5849601030349731,
0.3867762088775635,
0.7522372007369995,
0.44097432494163513
]
},
{
"class": "bowl",
"confidence": 96.1,
"box": [
0.6081124544143677,
0.18675625324249268,
0.6940791606903076,
0.27908602356910706
]
},
{
"class": "bowl",
"confidence": 92.54,
"box": [
0.60157710313797,
0.7178356647491455,
0.686405599117279,
0.7896746397018433
]
},
{
"class": "bowl",
"confidence": 90.81,
"box": [
0.7388676404953003,
0.6445244550704956,
0.8642051219940186,
0.8280567526817322
]
},
{
"class": "vase",
"confidence": 88.13,
"box": [
0.4349602460861206,
0.28476348519325256,
0.6799468398094177,
0.3464072346687317
]
},
{
"class": "bottle",
"confidence": 87.08,
"box": [
0.43529483675956726,
0.3596290946006775,
0.6646489500999451,
0.4041157364845276
]
},
{
"class": "person",
"confidence": 86.4,
"box": [
0.33264943957328796,
0.0,
1.0,
0.3069411516189575
]
},
{
"class": "dining table",
"confidence": 84.89,
"box": [
0.4721536934375763,
0.06169414892792702,
1.0,
0.9701008200645447
]
},
{
"class": "person",
"confidence": 81.4,
"box": [
0.1144622340798378,
0.37960779666900635,
0.605109691619873,
0.7577102780342102
]
},
{
"class": "fork",
"confidence": 73.61,
"box": [
0.741657018661499,
0.2172098457813263,
0.8201388120651245,
0.35427671670913696
]
},
{
"class": "bottle",
"confidence": 70.68,
"box": [
0.5082425475120544,
0.5854617953300476,
0.6571321487426758,
0.6402410864830017
]
},
{
"class": "chair",
"confidence": 66.25,
"box": [
0.4206991493701935,
0.7234641313552856,
0.6418891549110413,
0.8394129872322083
]
},
{
"class": "bottle",
"confidence": 64.52,
"box": [
0.44433778524398804,
0.29224613308906555,
0.6785905361175537,
0.34613561630249023
]
},
{
"class": "spoon",
"confidence": 52.35,
"box": [
0.741657018661499,
0.2172098457813263,
0.8201388120651245,
0.35427671670913696
]
},
{
"class": "person",
"confidence": 35.29,
"box": [
0.24801567196846008,
0.8235538601875305,
0.9413383603096008,
1.0
]
},
{
"class": "bottle",
"confidence": 30.45,
"box": [
0.40443146228790283,
0.44823896884918213,
0.6492757797241211,
0.5132496356964111
]
},
{
"class": "bowl",
"confidence": 28.11,
"box": [
0.668978214263916,
0.42829829454421997,
0.8111904263496399,
0.6795439720153809
]
}
]
},
{
"image": "Image2",
"detections": [
{
"class": "car",
"confidence": 99.69,
"box": [
0.5025718212127686,
0.7234153747558594,
0.7635517120361328,
1.0
]
},
{
"class": "person",
"confidence": 99.66,
"box": [
0.35710421204566956,
0.5226522088050842,
1.0,
0.7466851472854614
]
},
{
"class": "person",
"confidence": 97.54,
"box": [
0.3798726201057434,
0.2394254505634308,
1.0,
0.48451125621795654
]
},
{
"class": "car",
"confidence": 71.75,
"box": [
0.5769619941711426,
0.5101941823959351,
0.6027512550354004,
0.5349882245063782
]
},
{
"class": "car",
"confidence": 63.76,
"box": [
0.5723065733909607,
0.44900795817375183,
0.6150549054145813,
0.4942472577095032
]
},
{
"class": "car",
"confidence": 51.63,
"box": [
0.5714143514633179,
0.7081544399261475,
0.6126457452774048,
0.7267464995384216
]
},
{
"class": "bottle",
"confidence": 50.08,
"box": [
0.7275638580322266,
0.4005502760410309,
0.8659540414810181,
0.43661025166511536
]
},
{
"class": "bottle",
"confidence": 37.36,
"box": [
0.6906132698059082,
0.5630847215652466,
0.7971134185791016,
0.5943212509155273
]
},
{
"class": "truck",
"confidence": 29.39,
"box": [
0.5723065733909607,
0.44900795817375183,
0.6150549054145813,
0.4942472577095032
]
},
{
"class": "car",
"confidence": 27.56,
"box": [
0.57332843542099,
0.4903745949268341,
0.5998010635375977,
0.5043953061103821
]
}
]
}
]
}Image by image url (POST http://localhost:5050/image/by-url)
Request example
{
"images" : [
"https://images.unsplash.com/photo-1627662055794-94ab33f5913a?w=960"
]
}Response example
Image Upload (http://localhost:5050)
The output will be generated and saved into static/detections/OUTPUT_NAME.png
Request example
Response example
Reference
for more information