What is it?
The Trueface Visual Density Visionbox (TFVD) is a computer vision based crowd counting and analysis solution. TFVD is containerized, utilizes offline licensing, and features a simple REST API that can be used from any language, with no prior knowledge in computer vision or AI.
What can it do?
- 🧾 Count thousands of objects instantly from any perspective
- 🗾 Return a density map representing the spatial distribution of objects
How Does it Work?
TFVD utilizes artificial intelligence and computer vision to instantly count dense objects and analyze their spatial distribution. Returning a visual density and a crowd count estimate for the image. Simply post an image to your self hosted API to process it.
How to get started?
CPU
trueface/visual-density
GPU
trueface/visual-density-gpu
CPU
sudo docker run -e "TOKEN=$TOKEN" -v /data:/data -p 8080:8080 -d --name visual-density -it trueface/visual-densityGPU
sudo docker run --gpus all -e "TOKEN=$TOKEN" -v /data:/data -p 8080:8080 -d --name visual-density -it trueface/visual-density-gpuPath: /run
Method: POST
Payload
input_type: folder or payload.
type: image - reserved for future use.
input: input folder path in mounted folder or array of base64 encoded images. When passing a folder the API will process all images in the folder.
output: output folder path in mounted folder.
Payload Example
{
"input_type":"folder",
"type": "image",
"input": "/data/input",
"output": "/data/output"
}Sample Response
{
"data": [
{
"count": 139,
"density_heatmap": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAIBAQEBAQIBAQECAgICAgQDAgICAgUEBAMEBgUGBgYFBgYGBwkIBgcJBwYGCAsICQoKCgoKBggLDAsKDAkKCgr/wAALCABEAAMBAREA/8QAHwAAAQUBAQEBAQEAAAAAAAAAAAECAwQFBgcICQoL/8QAtRAAAgEDAwIEAwUFBAQAAAF9AQIDAAQRBRIhMUEGE1FhByJxFDKBkaEII0KxwRVS0fAkM2JyggkKFhcYGRolJicoKSo0NTY3ODk6Q0RFRkdISUpTVFVWV1hZWmNkZWZnaGlqc3R1dnd4eXqDhIWGh4iJipKTlJWWl5iZmqKjpKWmp6ipqrKztLW2t7i5usLDxMXGx8jJytLT1NXW19jZ2uHi4+Tl5ufo6erx8vP09fb3+Pn6/9oACAEBAAA/APxK/bSW3T9sX4spaMDEPiZrwiI7r/aE+P4V7f7I+g6U39sy1lsf2wPitZTKA8PxK11HAOQCNQnBp/7aX2s/tjfFk377p/8AhZmvec2c5f8AtCfJ/PNP/bdmjuf20Pi7cQtlJPihr7KfUHUZyKg/bJmiuP2vvirPAMI/xJ11kGOx1CcioP2tCx/ap+Jhbr/wsHWc/wDgdNVj9soQj9r74rC2OY/+Fk675Z/2f7Qnx+lL+2aYG/bB+K7Wy4jPxK13yx6L/aE+P0qH9rzj9rH4oDOf+Lia3/6XzV//2Q==",
"output": "results_e7cf8e0e-5275-46b1-b000-fb9f5c03ce77.json"
}
],
"message": "Predict count successfully!",
"status": "OK",
"statusCode": 200
}Path: /status
Method: GET
Sample Response
{
"data": null,
"message": "Your running TrueFace visual density",
"status": "OK",
"statusCode": 200
}Path: /shutdown
Method: GET
Sample Response:
{
"data": null,
"message": "exiting",
"status": "Accepted",
"statusCode": 202
}import base64
import requests
api_url = "http://localhost:8080"
headers = {
"content-type": "application/json"
}
data = {
"input_type":"payload",
"type": "image",
"input": [],
"output": "data/output",
}
image = "test.jpg"
data["input"].append(base64.b64encode(open("test.jpg", "rb").read()))
r = requests.post(api_url + "/run", headers=headers, json=data)
print(r.json())TOKENlicense token obtained from Trueface.WORKERS&THREADSTrueface utilizes Gunicorn. The number of Gunicorn workers and threads per worker can be arbitrarily set by the user using environment variables. Ex:
sudo docker run -e "WORKERS=2" -e "THREADS=10" -e "TOKEN=$TOKEN" -v $pwd/data:/data -p 8080:8080 -d --name visual-density -it trueface/visual-density- Trained and tested on images ranging from 10 to ~15,000 individuals. We recommend processing images with similar expected object counts.