Object counting

We are designing novel deep architectures for object counting under different training regimes: (i) counters based on trainable category-specific detectors and (ii) counters that do not require large training datasets but rely on using only few or no exemplars. The challenges are diverse scales of objects, densely poulated regions and appearance diversity.

Low-shot counting

The main goal of this research is development of computer-vision-based automated counters that do not require large training datasets, but are adapted to a previously unseen category by using only a few training examples (few-shot), no training examples (zero-shot) or text-based prompts (text-prompt-based).

Dense object counting in underwater imagery

The main goal of this research is development of computer-vision-based automated counters applicable underwater imagery. Such counters are crucial for processing extremely large datasets, vastly reducing the required manual labor and facilitating census orders of magnitude grater than what is possible with standard techniques. The methods leverage learning on specific type of images to maximize a task-specific detection performance.