Productive Crop Field Detection: A New Dataset and Deep Learning Benchmark Results

In precision agriculture, detecting productive crop fields is an essential practice that allows the farmer to evaluate operating performance separately and compare different seed varieties, pesticides, fertilizers. However, manually identifying often a time-consuming error-prone task. Previous studies explore methods detect using advanced machine learning algorithms, but they lack good quality labeled data. this context, we propose high-quality dataset generated by operation combined with Sentinel-2 images tracked over time. As far as know, it first one overcome of samples technique. sequence, apply semi-supervised classification unlabeled data state-of-the-art supervised self-supervised deep automatically. Finally, results demonstrate high accuracy in Positive Unlabeled learning, which perfectly fits problem where have confidence positive samples. Best performances been found Triplet Loss Siamese given existence accurate Contrastive Learning considering situations do not comprehensive available.