Benchmarking Dense and Indiscernible Object Counting with Blueberries

Real-world agricultural counting often operates in the extreme regime of \textbf{Dense and Indiscernible Object Counting (DIOC)}, where targets are tiny, clustered, and highly camouflaged. To facilitate research in this domain, we introduce \textbf{DIOCblueberry}, a large-scale benchmark that pushes the boundaries of visual perception. Unlike general datasets with salient objects, DIOCblueberry features extreme occlusion and camouflage. Compared to the popular FSC147 benchmark, it contains \textbf{1.9$\times$ more instances} per image (avg. 108) with an average box pixel ratio that is \textbf{7.9$\times$ smaller}, serving as a rigorous testbed for model robustness. Standard counting methods struggle in these scenarios due to severe visual ambiguity and scale mismatch. To address this, we propose \textbf{MaskCount}, a coarse-to-fine framework that incorporates semantic guidance. MaskCount leverages Vision-Language Models (CLIP) to generate pseudo segmentation masks for background suppression and employs a contrastive loss to maximize feature discriminability between fruits and foliage. Additionally, we design an edge-aware cropping mechanism to resolve boundary truncation in dense clusters. Extensive experiments demonstrate that MaskCount achieves a new state-of-the-art, reducing MAE and RMSE by \textbf{49.16\%} and \textbf{70.50\%} respectively on DIOCblueberry, with strong generalization to other agricultural scenes.

Counting fruit in real farm images is important for crop monitoring, yield estimation, and harvest planning, but it is often very difficult. Blueberries, for example, can be tiny, tightly packed together, partly hidden by leaves, and similar in color to their surroundings. To study this challenging problem, we introduce DIOCblueberry, a large benchmark dataset built from difficult real-world blueberry images. Compared with commonly used counting datasets, it contains more objects per image and much smaller fruits, making it a demanding test for AI counting systems. We also propose MaskCount, a new method designed to count fruits more reliably in these crowded and visually confusing scenes. MaskCount first helps the model focus less on background leaves and more on likely fruit regions. It then learns to better distinguish blueberries from surrounding foliage and carefully handles fruits near image edges that may be only partly visible. Experiments show that MaskCount greatly improves counting accuracy on DIOCblueberry and also performs well on other agricultural counting tasks. This work can support more reliable AI tools for precision agriculture.