International Journal For Multidisciplinary Research

E-ISSN: 2582-2160     Impact Factor: 9.24

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Call for Paper Volume 7, Issue 3 (May-June 2025) Submit your research before last 3 days of June to publish your research paper in the issue of May-June.

Object Detection in Low-Light Conditions Using YOLOV7

Author(s) Ms. Hujaimah Nida
Country India
Abstract Reduced contrast, increased noise, and shifting lighting levels are some of the factors that make object recognition in low light circumstances extremely difficult. Accurate identification in these situations is essential for applications where reliable performance is essential, such as driverless cars, security monitoring, and surveillance systems. This work makes use of the cutting-edge YOLOv7 object recognition model, It is renowned for its remarkable real-time processing speed and precision. The model is trained and evaluated using the ExDark dataset, which comprises twelve different object categories captured in different low-light scenarios.Adaptive histogram equalization and picture normalization are two preprocessing techniques used to enhance image quality and boost detection efficiency before to feeding the data into the YOLOv7 model. Training is done using pre-trained weights from the COCO dataset, which has been specially adapted to recognize objects in low light. The model is evaluated using standard assessment criteria, such as mean average precision (mAP), recall, and precision. With a recall of 0.6556 and a precision of 0.7153, YOLOv7 performs better on the same dataset than previous iterations such as YOLOv5 and IA-YOLO, according to the experimental results. The advanced network architecture of YOLOv7, which incorporates trainable Bag-of-Freebies and E-ELAN modules to enable efficient Feature extraction in low


light, is credited with the enhanced performance. The results highlight YOLOv7's potential for useful implementation in applications that need precise object detection in difficult illumination conditions. To further maximize real-time performance on edge devices, future improvements might include incorporating sophisticated picture augmentation techniques, image enhancement algorithms, and lightweight designs. The importance of continuously improving object detection technology is highlighted in this study in order to overcome the ongoing problems with low-light detection.
Keywords YOLOV7, deep learning, object detection, low-light video, ExDark dataset, Autonomous surveillance, Mean Average Precision (mAP).
Field Computer Applications
Published In Volume 7, Issue 3, May-June 2025
Published On 2025-05-04
DOI https://doi.org/10.36948/ijfmr.2025.v07i03.43710
Short DOI https://doi.org/g9hsht

Share this