Posts by Collection

portfolio

Portfolio item number 1

Short description of portfolio item number 1

Portfolio item number 2

Short description of portfolio item number 2

publications

HEVC based tampered video database development for forensic investigation

Published in Multimedia Tools and Applications, 2023

A dataset and analysis pipeline for tampered HEVC-encoded videos to support forensic investigation and benchmarking.

Recommended citation: Singla, N., Singh, J., Nagpal, S., & Tokas, B. (2023). HEVC based tampered video database development for forensic investigation. Multimedia Tools and Applications, 1–34.
Download Paper

Helmets Labeling Crops: Kenya Crop Type Dataset Created via Helmet-Mounted Cameras and Deep Learning

Published in Scientific Data, 2025

A data descriptor introducing a novel crop type dataset collected using helmet-mounted cameras and labeled using deep learning methods to support agricultural research in Kenya.

Recommended citation: Nakalembe, C., Zvonkov, I., Kerner, H., Frimpong, D., Mwangi, K., Kioko, J., Tokas, B., Jawanjal, K., Smith, I., Paliyam, A., & others (2025). Helmets Labeling Crops: Kenya Crop Type Dataset Created via Helmet-Mounted Cameras and Deep Learning. Scientific Data, 12(1), 1496.
Download Paper

Classification Drives Geographic Bias in Street Scene Segmentation

Published in CVPR, 2025

An analysis showing how classification decisions contribute to geographic biases in street scene segmentation models and recommendations to mitigate them.

Recommended citation: Nair, R., Tokas, B., Tseng, G., Rolf, E., & Kerner, H. (2025). Classification Drives Geographic Bias in Street Scene Segmentation. In Proceedings of the Computer Vision and Pattern Recognition Conference (pp. 629–638).
Download Paper

DPA: A one-stop metric to measure bias amplification in classification datasets

Published in NeurIPS, 2025

Most ML datasets today contain biases. When we train models on these datasets, they often not only learn these biases but can worsen them — a phenomenon known as bias amplification. We propose a better metric to measure bias amplification.

Recommended citation: Bhanu Tokas, Rahul Nair, & Hannah Kerner (2025). DPA: A one-stop metric to measure bias amplification in classification datasets. In The Thirty-ninth Annual Conference on Neural Information Processing Systems.
Download Paper

A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions

Published in WACV, 2026

Bias amplification for image captioning is different than classification. Impact of context of words in a sentence is missed by conventional bias amplification metrics. Metrics defined for image captioning are do not capture directionality and confound the source of biases. We propose a new metric that fixes these limitations.

Recommended citation: Nair, R., Tokas, B., & Kerner, H. (2026). A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) (pp. 255-264).
Download Paper

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.

Teaching experience 2

Workshop, University 1, Department, 2015