Distributed Data Parallel: Scaling AI Model Training
What truly makes for a better AI model? It’s a multifaceted journey, beginning with good data and winding through the intricacies of data...
my post
Beyond mAP: Better Evaluation of Bbox
In the pursuit of advancing computer vision technologies, evaluating the performance of instance segmentation or object detection models ...
Knowledge Distillation: A Comparative Exploration and the Role of Soft Targets
Knowledge Distillation is a process that refines large, complex models into smaller, efficient ones without losing much of the original p...
ViT vs CNN: A Technical Deep Dive for High-Resolution Vision Tasks
This isn’t just theory—it’s battle-tested. Over years of deploying vision systems for industrial-scale applications (from 4K surveillance...