CVDV: Computer Vision Data Visualization

A data visualization library for computer vision

Object Detection

Image Source: Traffic Signs Dataset in YOLO format

NOTE: OBJECT DETECTION DATA MUST BE YOLO FORMATTED

CVDV provides you the full analytics of your object detection dataset formatted according to the YOLO algorithm. The analysis performed using cvdv can help you in understanding the dataset. It can identify the class imbalance and bounding box size distribution according to different datasets.

---

1. Installation

---

git clone https://github.com/m3sibti/cvdv.git
cd cvdv
pip install -r requirements.txt

---

1. Running CVDV

python main.py --data_dir ./path/to/dataset --im_size XX

1.1 Parameters:

--data_dir: Path of dataset directory

--details_level: Levels of details you wannt to fetch
    . default: only class level information, or leave empty
    . all: for image level and object level information

--im_size: Size of the images for [SQUARE IMAGES]

--im_h: Height of the image for [NON-SQUARE IMAGES]
--im_w: Width of the image for [NON-SQUARE IMAGES]

---

2. Visualizations

Followings are the supported types of visualization in cvdv. The datset used for this analysis is available on Kaggle, Traffic Signs.

2.1 Class Distribution:

2.2 Bounding Box Pixel Histograms

Danger	Prohibitory

2.3 Bounding Box's Mean Size

2.4 Pixel's Color Co-relation

Danger	Prohibitory

---

Thank you for interest, Please provide your feedback.