PlantTrack: Task-Driven Plant Keypoint Tracking with Zero-Shot Sim2Real Transfer
This work addresses the challenge of plant feature tracking for agricultural tasks like phenotyping and harvesting, but it is incremental as it builds on existing foundational models and tracking methods.
The paper tackled the problem of tracking plant features in unstructured agricultural environments by proposing PlantTrack, which uses DINOv2 features and a keypoint predictor trained on synthetic data to enable zero-shot Sim2Real transfer, achieving effective tracking with as few as 20 synthetic training images.
Tracking plant features is crucial for various agricultural tasks like phenotyping, pruning, or harvesting, but the unstructured, cluttered, and deformable nature of plant environments makes it a challenging task. In this context, the recent advancements in foundational models show promise in addressing this challenge. In our work, we propose PlantTrack where we utilize DINOv2 which provides high-dimensional features, and train a keypoint heatmap predictor network to identify the locations of semantic features such as fruits and leaves which are then used as prompts for point tracking across video frames using TAPIR. We show that with as few as 20 synthetic images for training the keypoint predictor, we achieve zero-shot Sim2Real transfer, enabling effective tracking of plant features in real environments.