Comparison of Single-Shot and Two-Shot Deep Neural Network Models for Whitefly Detection in IoT Web Application

Abstract

In this study, we have compared YOLOv4, a single-shot detector to Faster-RCNN, a two-shot detector to detect and classify whiteflies on yellow-sticky tape (YST). An IoT remote whitefly monitoring station was developed and placed in a whitefly rearing room. Images of whiteflies attracted to the trap were recorded 2× per day. A total of 120 whitefly images were labeled using labeling software and split into a training and testing dataset, and 18 additional yellow-stick tape images were labeled with false positives to increase the model accuracy from remote whitefly monitors in the field that created false positives due to water beads and reflective light on the tape after rain. The two-shot detection model has two stages: region proposal and then classification of those regions and refinement of the location prediction. Single-shot detection skips the region proposal stage and yields final localization and content prediction at once. Because of this difference, YOLOv4 is faster but less accurate than Faster-RCNN. From the results of our study, it is clear that Faster-RCNN (precision—95.08%, F-1 Score—0.96, recall—98.69%) achieved a higher level of performance than YOLOv4 (precision—71.77%, F-1 score—0.83, recall—73.31%), and will be adopted for further development of the monitoring station.

A Review of Successes and Impeding Challenges of IoT-Based Insect Pest Detection Systems for Estimating Agroecosystem Health and Productivity of Cotton

“…Boll weevil, Cotton aphid, Cotton bollworm (larva), Cotton bollworm (adult), tobacco budworm (larva), Tobacco budworm S (adult), Soybean looper, Fall armyworm (larva), Fall armyworm (adult), Cotton leafworm, Cotton whitefly, Cotton bug, Pink bollworm, southern armyworm, and red spider mite [1] Cotton aphids, Flea beetles, Flax budworms, and Red spider mites [11] Mexican cotton boll weevil, Fall armyworm, Cotton bollworm, Cotton aphid, Cotton whitefly, Green stink bug, Neotropical brown stink bug, Soybean looper [14] Assassin Bug, Three-Corned Alfalfa Hopper, and Convergent lady beetle [21] Red spider mites and Leaf miner [27] Pink and American bollworms [28,29] American bollworm, Ash weevil, Blossom thrips, Brown cotton moth, Brown soft scale, Brown-spotted locust, Cotton aphid, Cotton leaf roller, Cotton leafhopper, Cotton looper, Cotton stem weevil, Cotton whitefly, Cream drab, Cutworm, Darth maul moth imago, Darth maul moth, Desert locust, Dusky cotton bug, Giant red bug, Golden twin spot tomato looper, Green stink bug, Grey mealybug, Hermolaus, Latania scale, Madeira mealybug, Mango mealybug, Megapulvinaria, Cotton stainers, Menida, Menida-versicolor, White-spotted flea beetle, Myllocerus-subfasciatus, Sri Lankan weevil, Painted bug, Pink bollworm, Brown-winged green bug, Poppiocapsidea, Red-banded shield bug, Red cotton bug, Red hairy caterpillar, Solenopsis mealybug, Spherical mealybug, Spotted bollworm imago, Spotted bollworm, Tobacco caterpillar Tomentosa, Transverse moth, Tussock caterpillar, and Yellow cotton scale [29] Cotton whitefly [1,14,23,[29][30][31][32]…”

“…In a similar study, 120 images were used to compare the performance of Fast R-CNN and YOLOv4 detection models [32]. The Faster-RCNN model had a better accuracy of 95.08% (F-1 Score: 0.96, recall: 98.69%) compared to a YOLOv4 model that indicated a lower accuracy of 71.77% (F-1 score: 0.83, recall: 73.31%).…”

“…For advanced pest tracking systems, the platform should provide multiple sensor data and work in two ways: it should deliver field information to the user and allow the user to query it at any moment, as illustrated in Figure 2. In cotton fields, the systems to detect flying insects in cotton fields named the "selfcleaning trap" (Figure 3a) and "remote whitefly monitor" (Figure 3b) were shown in [32] and [45], respectively. The "remote whitefly monitor" and "self-cleaning trap" use sticky traps to fix the insects to the camera sensor's position.…”

“…These devices can also potentially include models that reasonably forecast pest dynamics, including weather conditions. In cotton fields, the systems to detect flying insects in cotton fields named the "selfcleaning trap" (Figure 3a) and "remote whitefly monitor" (Figure 3b) were shown in [32] and [45], respectively. The "remote whitefly monitor" and "self-cleaning trap" use sticky traps to fix the insects to the camera sensor's position.…”

Detection of whitefly pests in crops employing image enhancement and machine learning

“…The accuracy obtained was 62%, 73.7% and 62.5%. [9] Deep learning-based approach for whitefly detection on a yellow-sticky tape (YST) at an IoT remote whitefly monitoring station. Precision of 95.08%, F-1 Score of 0.96 and recall of 98.69%, beating YOLOv4 which obtained Precision of 71.77%, F-1 Score of 0.83 and recall of 73.31%,…”