Sameer Ambekar

Deep learning | Computer vision

I am pursuing my Ph.D. (Doctoral Researcher) at Technical University of Munich (TU Munich) and Helmholtz Munich in Deep learning under Prof. Dr. Julia Schnabel, Dr. Daniel Lang and mentored by Prof. Dr. Stefan Bauer. My academic journey includes Masters in Artificial Intelligence (MSc AI) at the University of Amsterdam (UvA) and also worked as Research Intern at the university research AIM lab, UvA. For MSc AI thesis, I addressed Test-Time Adaptation of classifiers for Domain Generalization utilizing Meta learning, Variational inference for efficient multi-source training and transferrable features under Prof. dr. Cees Snoek, Prof. Xiantong Zhen and Zehao Xiao.

Before MSc AI, I worked as Research Assistant (RA) at IIT Delhi India under Prof. Prathosh A.P. in domain adaptation through Generative Latent search in the VAE latent space. Before working as RA at IITD, I worked at Indian Council of Medical Research (ICMR) as a Researcher under Dr. Subarna Roy (Scientist G), and Mr. Pramod Kumar (Scientist C).

Additionally, I also serve as a Reviewer at top-tier conferences and journals such as NeurIPS, CVPR, ICML, ECCV, ICCV, WACV, IEEE Transactions on Neural Networks and Learning Systems, Elsevier Applied soft computing. Furthermore, serve as a Mentor for Neuromatch deep learning course.

I'm always interested in new collaborations, so feel free to email me to chat!

I am interested in Test-time adaptation and Unsupervised learning.
Utilizing learned representations and patterns to transfer to unseen domains although high-dimensional and streaming batches of data with distribution shifts.

My efforts have focussed on addressing problems in Test-time adaptation, Domain Generalization, Domain Adaptation through Variational inference, meta learning, surrogate model updates and model predictions.

We consider the problem of test-time domain generalization, where a model is trained on several source domains and adjusted on target domains never seen during training. Different from the common methods that fine-tune the model or adjust the classifier parameters online, we propose to generate multiple layer parameters on the fly during inference by a lightweight meta-learned transformer, which we call GeneralizeFormer. The layer-wise parameters are generated per target batch without fine-tuning or online adjustment. By doing so, our method is more effective in dynamic scenarios with multiple target distributions and also avoids forgetting valuable source distribution characteristics. Moreover, by considering layer-wise gradients, the proposed method adapts itself to various distribution shifts. To reduce the computational and time cost, we fix the convolutional parameters while only generating parameters of the Batch Normalization layers and the linear classifier. Experiments on six widely used domain generalization datasets demonstrate the benefits and abilities of the proposed method to efficiently handle various distribution shifts, generalize in dynamic scenarios, and avoid forgetting.

First, we propose probabilistic pseudo-labeling of target samples to generalize the source-trained model to the target domain at test time. We formulate the generalization at test time as a variational inference problem by modeling pseudo labels as distributions to consider the uncertainty during generalization and alleviate the misleading signal of inaccurate pseudo labels. Second, we learn variational neighbor labels that incorporate the information of neighboring target samples to generate more robust pseudo labels. Third, to learn the ability to incorporate more representative target information and generate more precise and robust variational neighbor labels, we introduce a meta-generalization stage during training to simulate the generalization procedure.

We cast the skin segmentation problem as that of target-independent Unsupervised Domain Adaptation (UDA) where we use the data from the Red-channel of the visible-range to develop skin segmentation algorithm on NIR images. We propose a method for target-independent segmentation where the ‘nearest-clone’ of a target image in the source domain is searched and used as a proxy in the segmentation network trained only on the source domain. We prove the existence of ‘nearest-clone’ and propose a method to find it through an optimization algorithm over the latent space of a Deep generative model based on variational inference.

We propose a novel work named SKDCGN that attempts knowledge transfer using Knowledge Distillation (KD). In our proposed architecture, each independent mechanism (shape, texture, background) is represented by a student 'TinyGAN' that learns from the pretrained teacher 'BigGAN'. We demonstrate the efficacy of the proposed method using state-of-the-art datasets such as ImageNet, and MNIST by using KD and appropriate loss functions. Moreover, as an additional contribution, our paper conducts a thorough study on the composition mechanism of the CGNs, to gain a better understanding of how each mechanism influences the classification accuracy of an invariant classifier.

Academic Project, MSc AI, UvA.

We introduce a state-of-the-art technique, termed as Twin Augmentation, for modifying popular pre-trained deep learning models. Twin Augmentation boosts the performance of a pre-trained deep neural network without requiring re-training. Experiments show, that across a multitude of classifiers, Twin Augmentation is very effective in boosting the performance of given pre-trained model for classification in imbalanced settings.

I pursued my MSc thesis (48ECTS, Grade: Excellent) to address Test-time Adaptation for Domain Generalization by proposing 2 methods : (i) Formulating it as a Variational Inference problem and meta-learn Variational neighbor labels in a probabilistic framework alongside exploring neighborhood information (ii) Surrogate update of the model without backpropagation.
Courses enrolled: Curriculum based and Research based.

Last updated: