AgriCountDINO improves plant counting with fewer parameters
AgriCountDINO: Parameter-Efficient Exemplar-Guided Counting and Localization in Agriculture
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Counting and locating plants correctly helps farmers understand crop health and yields. The authors created AgriCountDINO, a method that uses example images of plants to find and count them without needing to retrain for every new plant type. This method uses fewer parameters than previous models and can detect plants more accurately while also giving their locations. It works well even on new kinds of objects it hasn’t seen before, showing flexibility.
What this means in practice
- •For agriculture technology developers: Integrate efficient plant counting and localization to improve phenotyping and yield estimation tools without extensive retraining.
- •For wildlife monitoring teams: Use exemplar-based counting methods to identify and locate individual animals or plants in new environments without retraining models.
Authors
Shengjie Guo, Xin Li, Borjana Arsova, Hanno Scharr, Silvio Salvi
Abstract
Accurate counting and localization of plants and their organs support phenotyping and yield estimation, yet target appearance, scale, and density vary widely across species and imaging conditions. Exemplar boxes specify the target without category-specific retraining, and point predictions identify the individual instances contributing to the count. We introduce AgriCountDINO, a parameter-efficient exemplar-guided framework for joint counting and localization. It conditions frozen multiscale DINOv3 features on exemplar appearance and size, then progressively decodes them into target points. Missed-object recovery extends supervision to targets overlooked by initial matching, and exemplar-adaptive point NMS filters duplicate predictions according to exemplar scale. With 8.4M trainable parameters, approximately one-tenth of TasselNetV4's, AgriCountDINO achieves a three-shot MAE of 11.92 on the TPC-268 benchmark, reducing counting error by 9.7\% while providing individual target locations. Trained only on TPC-268, it achieves a zero-shot MAE of 14.25 on unseen generic object categories in FSC-147, improving upon the best compared zero-shot method by 6.0\% without target-domain training or fine-tuning.