Video-Based Sexual Harassment Detection Using Fine-Tuned Video Transformers

Jubayer- Al- Mahmud; Md Shamimul Islam; Md Nur Hossain

cs.CVOct 11, 2026

Video-Based Sexual Harassment Detection Using Fine-Tuned Video Transformers

Authors: Jubayer- Al- Mahmud, Md Shamimul Islam, Md Nur Hossain

Organizations: Department of CSE, Manarat International University, Dhaka, Bangladesh · Department of Computer Science, Florida Polytechnic University, Lakeland, Florida, USA

Abstract

Sexual harassment represents a pervasive societal challenge, with a substantial proportion of incidents remaining unreported due to insufficient evidence and delayed intervention mechanisms. Smart surveillance systems capable of detecting such incidents in real-time could assist law enforcement and serve as a deterrent. However, video-based harassment detection poses significant challenges, including variations in clothing, illumina- tion, partial occlusions, low resolution, and diverse camera view- points. Despite its social importance, this problem has received limited research attention, primarily due to the scarcity of large- scale annotated video datasets. To address this gap, we extend the Sexual Harassment Video (SHV) dataset to 600 short video clips sourced from YouTube and annotated into harassment and non-harassment categories using consistent labeling guidelines. We fine-tune three state-of-the-art video transformer models, VideoMAE, ViViT, and TimeSformer, pretrained on Kinetics- 400, for binary harassment classification. Experimental results show that all three models achieve competitive classification performance, with TimeSformer attaining the highest test accu- racy of 92.27%, outperforming VideoMAE 91.71% and ViViT 91.16% on the extended SHV test set. Our analysis reveals that performance is notably influenced by dataset size, video quality, frame sampling strategy, and model architecture. This work presents a comprehensive evaluation of modern video transformers for this sensitive real-world task.

Figures & tables

Fig. 1: Overview of the data preparation and model initialization pipeline for sexual harassment video classification.

Fig. 2: Example of 16 uniformly sampled frames extracted from a harassment video clip in the SHV dataset.

Description	Details
Total Classes	2
Total Videos	600
Harassment Videos	282
Non-Harassment Videos	318
Training Set (Harassment)	197
Training Set (Non-Harassment)	222

TABLE I: Sexual Harassment Video (SHV) Dataset Statistics

Configuration	VideoMAE	ViViT	TimeSformer
Pretrained Checkpoint	MCG-NJU/videomae-base-finetuned-kinetics	google/vivit-b-16x2-kinetics400	facebook/timesformer-base-finetuned-k400
Image Size	224	224	224
Num. Frames	16	32	8
Patch Size	16	16 $\times$ 2	16
Hidden Size	768	768	768
Num. Attention Heads	12	12	12

TABLE II: Comparison of video transformer model architectures and their configurations used in this study.

Fig. 3: Architecture overview of the three video transformer models used in this study: (a) VideoMAE, (b) ViViT, and (c) TimeSformer.

Model	Class	Test Loss	TP	FP	FN	TN	Precision	Recall	F1-Score	Macro F1
VideoMAE	Harassment	0.2601	88	8	7	78	0.907	0.918	0.912	0.917
	Non-Harassment		78	7	8	88	0.926	0.917	0.921
ViViT	Harassment	0.2474	93	3	13	72	0.960	0.847	0.900	0.910
	Non-Harassment		72	13	3	93	0.877	0.969	0.921
TimeSformer	Harassment	0.2079	94	2	12	73	0.973	0.859	0.912	0.922
	Non-Harassment		73	12	2	94	0.887	0.979	0.931

TABLE III: Per-Class and Average Performance of Fine-Tuned Video Transformer Models on the SHV Test Set.

Fig. 4: Test loss and accuracy comparison.

Hyperparameter	Value
Batch Size	2
Max Epochs	20
Initial Learning Rate	$10^{-5}$
Weight Decay	$10^{-4}$
Optimizer	AdamW
Loss Function	Cross-Entropy Loss

TABLE IV: Training Hyperparameters for Fine-Tuning

Model	Accuracy (%)
VideoMAE	91.71
ViViT	91.16
TimeSformer	92.27
CNN-LSTM [ 1 ]	70

TABLE V: Accuracy comparison on the SHV test set.

Video Content Moderation

Video content moderation detects and explains harmful, misleading, or policy-violating material in videos, including its visual, audio, and temporal context.

See all 23 papers

Apr 1, 2026

HarassGuard: Detecting Harassment Behaviors in Social Virtual Reality with Vision-Language Models

Social Virtual Reality (VR) platforms provide immersive social experiences but also expose users to serious risks of online harassment. Existing safety measures are largely reactive, while proactive solutions that detect harassment behavior during an incident often depend on sensitive biometric data, raising privacy concerns. In this paper, we present HarassGuard, a vision-language model (VLM) based system that detects physical harassment in social VR using only visual input. We construct an IRB-approved harassment vision dataset, apply prompt engineering, and fine-tune VLMs to detect harassment behavior by considering contextual information in social VR. Experimental results demonstrate that HarassGuard achieves competitive performance compared to state-of-the-art baselines (i.e., LSTM/CNN, Transformer), reaching an accuracy of up to 88.09% in binary classification and 68.85% in multi-class classification. Notably, HarassGuard matches these baselines while using significantly fewer fine-tuning samples (200 vs. 1,115), offering unique advantages in contextual reasoning and privacy-preserving detection.

Junhee Lee, Minseok Kim, Hwanjo Heo +2

Kwangwoon University · ETRI · Chungbuk National University

Virtual Reality

Ask PDF

May 23, 2025

FuseMamba-VD: Dual Branch VideoMamba with Gated Class Token Fusion for Violence Detection

The rapid proliferation of surveillance cameras has increased the demand for automated violence detection. While CNNs and Transformers have shown success in extracting spatio-temporal features, they struggle with long-term dependencies and computational efficiency. We propose FuseMamba-VD: Dual Branch VideoMamba with Gated Class Token Fusion (GCTF), an efficient architecture combining a dual-branch design and a state-space model (SSM) backbone where one branch captures spatial features, while the other focuses on temporal dynamics. The model performs continuous fusion via a gating mechanism from the spatial branch into the temporal branch to enhance detection of violent activities even in challenging surveillance scenarios. We also present a new benchmark by merging RWF-2000, RLVS, SURV and VioPeru datasets in video violence detection, ensuring strict separation between training and testing sets. Experimental results demonstrate that our model achieves state-of-the-art performance on this benchmark and also on DVD dataset which is a recently introduced dataset on video violence detection, offering an optimal balance between accuracy and computational efficiency, demonstrating the promise of SSMs for scalable, resource efficient video violence detection. The code and pre-trained models are available at https://github.com/damith92/FuseMamba-VD.

Damith Chamalke Senadeera, Muhammad Awais, Shibo Li +2

Queen Mary University of London, UK · Digital Environment Research Institute (DERI), London, UK · Queen’s University Belfast, UK

Video Surveillance

Ask PDF

Aug 16, 2026

CLARA: Clip-Level Multimodal Alignment with VLM-Derived Rationales for Hateful Video Detection

Hateful video detection has become increasingly important with the rapid growth of video-centric social media platforms, given the serious risks that hate speech poses to both individual well-being and social cohesion. Compared with text or static multimodal content, hateful video detection remains underexplored and significantly more challenging, as hateful meaning often arises from complex interactions among multimodal cues, including speech, audio, and visual content. Moreover, such signals are often brief, implicit, and temporally dependent, making them difficult to capture using conventional video-level representations. In this work, we propose CLARA, a clip-level multimodal framework for hateful video detection. Instead of treating a video as a single instance, CLARA models it as a sequence of fine-grained clips, enabling more precise capture of temporally localized hateful signals. We introduce a Mixture-of-Experts clip encoder for adaptive multimodal alignment, a local-global segment contrastive objective to jointly model short-term cues and long-range temporal dependencies, and VLM-derived rationales integrated via a gated Transformer to provide high-level semantic guidance. Extensive experiments on three hateful video datasets demonstrate that CLARA consistently outperforms state-of-the-art methods. Further ablation studies and parameter analyses validate the effectiveness of each component.

Yuchen Zhang, Shuang Dai, Zeyu Fu +3

Institute for Analytics and Data Science, University of Essex, Colchester, United Kingdom · Department of Engineering, University of Exeter, Exeter, United Kingdom · Department of Computer Science, University of Exeter, Exeter, United Kingdom +1

Hate Speech Detection Video Content Moderation Temporal Video Understanding

Ask PDF

Abstract

Video Content Moderation

Explore similar work

HarassGuard: Detecting Harassment Behaviors in Social Virtual Reality with Vision-Language Models

FuseMamba-VD: Dual Branch VideoMamba with Gated Class Token Fusion for Violence Detection

CLARA: Clip-Level Multimodal Alignment with VLM-Derived Rationales for Hateful Video Detection