Search

The rise of convolutional networks in computer vision, especially for generic object detection, has led to the emergence of a myriad of efficient and precise object detection models. Typically, deep learning-driven object detectors operate in two phases: initially, they utilize convolutional networks to extract compact feature embeddings from images; subsequently, these embeddings are used to pinpoint localized object positions. Rooted in convolutional networks, these generic object detection models have the capability to learn from vast datasets that comprise hundreds of thousands of images with thousands of objects. This vast data training gives them unparalleled generalization capabilities, setting them apart from traditional methods. With the swift pace of research, new object detection models are frequently unveiled, each striving for state-of-the-art performance on renowned benchmarks. Given the abundance of viable models, selecting the optimal one can be a daunting task. In this paper, we offer a succinct overview of widely recognized object detectors, emphasizing their architectural distinctions, and presenting a quantitative comparison in terms of accuracy and inference speeds using the popular 2017 Common Objects in Context dataset.

We present an end-to-end smart harvesting solution for precision agriculture. Our proposed pipeline begins with yield estimation that is done through the use of object detection and tracking to count fruit within a video. We use and train You Only Look Once model (YOLO) on video clips of apples, oranges and pumpkins. The bounding boxes obtained through objection detection are used as an input to our selected tracking model, DeepSORT. The original version of DeepSORT is unusable with fruit data, as the appearance feature extractor only works with people. We implement ResNet as DeepSORT’s new feature extractor, which is lightweight, accurate and generically works on different fruits. Our yield estimation module shows accuracy between 91–95% on real footage of apple trees. Our modification successfully works for counting oranges and pumpkins, with an accuracy of 79% and 93.9% with no need for training. Our framework additionally includes a visualization of the yield. This is done through the incorporation of geospatial data. We also propose a mechanism to annotate a set of frames with a respective GPS coordinate. During counting, the count within the set of frames and the matching GPS coordinate are recorded, which we then visualize on a map. We leverage this information to propose an optimal container placement solution. Our proposed solution involves minimizing the number of containers to place across the field before harvest, based on a set of constraints. This acts as a decision support system for the farmer to make efficient plans for logistics, such as labor, equipment and gathering paths before harvest. Our work serves as a blueprint for future agriculture decision support systems that can aid in many other aspects of farming.