Person Computer Vision Dataset
About Person Dataset
This project develops a robust computer vision model to estimate crowd density from live urban video feeds, with applications in public safety, event management, urban planning, and retail analytics.
Dataset:
522 images from New York and Tokyo live feeds.
Captures diverse conditions: lighting, weather, times of day, and varying crowd densities.
Challenges include heavy occlusion, scale variation, and variable lighting.
Annotation Strategy:
Point-based annotation: each person labeled with a single point on their head/torso.
Annotated using Roboflow for efficient, accurate labeling.
Model Architecture:
Base: VGG16 pre-trained on ImageNet, used as a frozen feature extractor.
Custom regression layers added on top to predict total crowd count per image.
Regression approach is chosen over detection to handle high occlusion.
Goal & Applications:
Accurate real-time crowd density estimation.
Supports public safety alerts, traffic/pedestrian management, event planning, and retail analytics.
Use Free Person Detection API
Powered by general detection model
pip install inference-sdk# 1. Import the library
from inference_sdk import InferenceHTTPClient
# 2. Connect to your workspace
client = InferenceHTTPClient(
api_url="https://serverless.roboflow.com",
api_key="API_KEY"
)
# 3. Run your workflow on an image
result = client.run_workflow(
workspace_name="<YOUR_WORKSPACE>",
workflow_id="<YOUR_WORKFLOW_ID>",
images={
"image": "YOUR_IMAGE.jpg" # Path to your image file
},
parameters={
"classes": "Person"
},
use_cache=True # cache workflow definition for 15 minutes
)
# 4. Get your results
print(result)Run on custom image
Drop an image here or click to upload
Tell the agent what you want to build.
Cite This Project
LicenseCC BY 4.0If you use this dataset in a research paper, please cite it using the following BibTeX:
@misc{ person-j8mjg-ctb49_dataset,
title = { Person Dataset },
type = { Open Source Dataset },
author = { Shadiq Workspace },
howpublished = { \url{ https://universe.roboflow.com/shadiq-workspace/person-j8mjg-ctb49 } },
url = { https://universe.roboflow.com/shadiq-workspace/person-j8mjg-ctb49 },
journal = { Roboflow Universe },
publisher = { Roboflow },
year = { 2026 },
month = { apr },
note = { visited on 2026-07-29 },
}