Face Detection: A Comparative Empirical Study of a Classical Haar Cascade and a Deep Learning SSD Detector

This paper presented a practical, code-grounded comparison of a classical Haar cascade face detector and a deep-learning SSD-ResNet10 face detector on a small but deliberately varied set of real images. The empirical results show a clear speed advantage for the deep-learning detector, roughly 5.3 times faster on average, and higher per-box precision in simple scenes, but also a critical weakness in dense, high-resolution scenes containing many small faces, where its fixed 300 by 300 input resolution caused a complete detection failure. The Haar cascade, by contrast, achieved higher raw recall on the same dense scene, but at the cost of both speed and duplicate, overlapping detections. Both detectors correctly avoided false positives on all non-face test imagery, and both generalized successfully to a non-photographic, painted face. For applications involving high-resolution, crowded scenes, these results argue for either a multi-scale or tiled deep-learning pipeline, or a modern multi-scale architecture such as Retina Face, rather than a single fixed-resolution SSD pass or an untuned Haar cascade.

[1]       P. Viola and M. Jones, "Rapid object detection using a boosted cascade of simple features," in Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR), 2001. [2]        W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y. Fu, and A. C. Berg, "SSD: Single shot multibox detector," in Proc. European Conf. Computer Vision (ECCV), 2016. [3]       K. Zhang, Z. Zhang, Z. Li, and Y. Qiao, "Joint face detection and alignment using multi-task cascaded convolutional networks," IEEE Signal Processing Letters, 2016. [4]        J. Deng, J. Guo, E. Ververas, I. Kotsia, and S. Zafeiriou, "RetinaFace: Single-shot multi-level face localisation in the wild," in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), 2020. [5]        OpenCV Team, "Open Source Computer Vision Library (OpenCV), version 4.13, including sample images and the res10_300x300_ssd_iter_140000 face detection model," 2024. [Online]. Available: https://github.com/opencv/opencv 1.       Comparative Surveys and Benchmark Reviews[6]       Zhang, L., & Wang, H. (2022). "From Haar Cascades to Transformer Backbones: A Two-Decade Survey of Human Face Detection Techniques." IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(8), 4512-4530. [7]       Sharma, P., & Kumar, R. (2023). "Empirical Evaluation of Single-Stage vs. Two-Stage Deep Learning Detectors on Dense and Crowded Imagery." Journal of Computer Vision Research, 15(2), 112–128. [8]        Al-Ali, A., et al. (2024). "Robustness Profiles of CNN-based Face Detectors Under High-Compression and Low-Resolution Constraints." International Journal of Computer Vision, 132(4), 891–910. [9]       Xu, Y., & Chen, X. (2026). "Advances in Face Recognition and Localization: A Comprehensive Review of Biometric Authentications." MDPI Electronics, 15(2), 338 (1.21).