Crop Disease Diagnosis with Deep Learning-Based Image Captioning and Object Detection

The number of people participating in urban farming and its market size have been increasing recently. However, the technologies that assist novice farmers are still limited. There several previously researched deep learning-based crop disease diagnosis solutions. these techniques only focus on CNN-based detection do not explain characteristics symptoms based severity. In order to prevent spread diseases crops, it is important identify advance cope with them as soon possible. Therefore, we propose an improved solution which can give practical help farmers. proposed consists two representative methods: Image Captioning Object Detection. model describes prominent disease, according severity detail, by generating diagnostic sentences grammatically correct semantically comprehensible, along presenting accurate name it. Meanwhile, Detection detects infected area recognize part damaged assure accuracy sentence generated model. employs InceptionV3 encoder Transformer a decoder, while YOLOv5 average BLEU score 64.96%, be considered high performance generation and, meanwhile, mAP50 for 0.382, requires further improvement. Those results indicate allows precise elaborate information diseases, thereby overall reliability diagnosis.