Code for "Complex-Valued Depthwise Separable Convolutional Neural Network for Automatic Modulation Classification"

LUPI-OD - A novel methodology to improve object detection accuracy without increasing model size or complexity. | 2025 European Workshop on Visual Information Processing (EUVIP) | M.Sc. ICT (By Research) Dissertation | University of Malta

Overview of self-supervised learning of tiny models, including distillation-based methods (aks. self-supervised distillation) and non-distillation methods.

An awesome list of "small but mighty" models and resources.

Enhancing lightweight models for efficient sensor-based human activity recognition

Notes and resources from Qualcomm On-device AI course, provided by DeepLearningAI

Official PyTorch code for the paper "Towards Faster and Efficient Lightweight Image Super Resolution using SwinV2 Transformers and Fourier Convolutions." PyTorch · SwinV2 Transformers · Fast Fourier Convolutions · TorchServe · Docker. 33.55% fewer params, up to 60% faster than SwinIR.

A comprehensive study on optimizing hyperparameters for lightweight deep learning models aimed at real-time image classification on edge devices. Includes training configs, scripts, and performance benchmarks.

[Research] Lightweight, boundary-aware, noise-robust polyp segmentation (Kvasir-SEG & CVC-ClinicDB) with calibration and ECE maps

Fake detection using BERT & smooth pruning

Lightweight Big Rock segmentation on the AI4Mars dataset under space-grade hardware and radiation constraints, with a simulated fault-injection survivability stack.