no code implementations • 23 Feb 2024 • Purbayan Kar, Vishal Chudasama, Naoyuki Onoe, Pankaj Wasnik, Vineeth Balasubramanian
To effectively utilize the newly proposed augmentation technique, we employ a Siamese architecture-based training mechanism with a Deep Canonical Correlation Analysis (DCCA)-based loss to achieve collective learning of high-level feature representations from two different views of the input images.
Ranked #1 on Facial Landmark Detection on WFLW
no code implementations • 4 Jun 2023 • Purbayan Kar, Vishal Chudasama, Naoyuki Onoe, Pankaj Wasnik
Semi-supervised object detection (SSOD) has made significant progress with the development of pseudo-label-based end-to-end methods.
no code implementations • 5 Jun 2022 • Vishal Chudasama, Purbayan Kar, Ashish Gudmalwar, Nirmesh Shah, Pankaj Wasnik, Naoyuki Onoe
We introduce a new feature extractor to extract latent features from the audio and visual modality.
Ranked #10 on Emotion Recognition in Conversation on MELD
no code implementations • 28 Mar 2022 • Hari Shankar, Adithya Narayan, Shefali Jain, Divya Singh, Pooja Vyas, Nivedita Hegde, Purbayan Kar, Abhi Lad, Jens Thang, Jagruthi Atada, Duy Nguyen, PS Roopa, Akhila Vasudeva, Prathima Radhakrishnan, Sripad Krishna Devalla
Multiple studies have demonstrated that obtaining standardized fetal brain biometry from mid-trimester ultrasonography (USG) examination is key for the reliable assessment of fetal neurodevelopment and the screening of central nervous system (CNS) anomalies.