Deeptrace: South Asia Deepfake Detection Challenge
Embedded Files
Mission of the Project
Mission of the Project
The South Asia Deepfake Detection Challenge is a multi-country research initiative to build culturally grounded datasets, benchmarks, and detection tools for AI-generated media in Sri Lanka, India, Pakistan, Bangladesh, and Nepal. Our goal is to advance deepfake detection that truly reflects the region’s languages, aesthetics, and socio-political realities.
Objectives of the South Asia Deepfake Detection Challenge
1. Build a culturally representative multimodal deepfake dataset
1. Build a culturally representative multimodal deepfake dataset
Collect image, video, audio, and text-based deepfakes that reflect South Asian languages, cultural cues, political contexts, and real-world manipulation styles.
2. Develop benchmark tasks and evaluation protocols for South Asian deepfakes
2. Develop benchmark tasks and evaluation protocols for South Asian deepfakes
Design region-aware benchmark tests that capture local challenges—such as low-quality media, mobile-edited fakes, linguistic code-mixing, and culturally specific manipulations.
3. Train and compare deepfake detection models optimized for regional content
3. Train and compare deepfake detection models optimized for regional content
Develop machine-learning models that can detect deepfakes across diverse South Asian languages, faces, and contexts, and evaluate them against existing global tools.
4. Build user-facing tools and applications for journalists and the public
4. Build user-facing tools and applications for journalists and the public
Create accessible interfaces—browser tools, mobile prototypes, and fact-checking utilities—that help people identify, contextualize, and report AI-generated media.
Questions?
Questions?
Contact [email] to get more information on the project
Page updated
Google Sites
Report abuse