Benchmark datasets for detection and identification of insects from camera trap images with deep learning

<strong>Insect benchmark datasets for training, validation and test (train1201.zip, val1201.zip and test1201.zip) with time-lapse images as described in paper:</strong> Bjerge K, Alison J, Dyrmann M, Frigaard C.E., Mann H. M. R., Høye T.T., Accurate detection and identification of insects from camera trap images with deep learning, bioRxiv:10.1101/2022.10.25.513484v1 Labels in <strong>YOLO format: ultralytics/yolov5: label format</strong> The annotated training and validation datasets contains insects of nine different species as listed below: 0 <em>Coccinellidae septempunctata</em> 1 <em>Apis mellifera</em> 2 <em>Bombus lapidarius</em> 3 <em>Bombus terrestris</em> 4 <em>Eupeodes corolla</em> 5 <em>Episyrphus balteatus</em> 6 <em>Aglais urticae</em> 7 <em>Vespula vulgaris</em> 8 <em>Eristalis tenax</em> The test dataset contains additional classes of insects. 9 Non-Bombus Anthophila 10 Bombus spp. 11 Syrphidae 12 Fly spp. 13 Unclear insect 14 Mixed animals:<br> ——————————<br> Rhopalocera<br> Non-Anthophila Hymenoptera<br> Non-Syrphidae Diptera<br> Non-Conccinalidae Coleoptera<br> Concinellidae<br> Other animals <strong>There are two naming conventions for image (.jpg) and label (.txt) files.</strong> <em>Background images without insects are named</em>:<br> “<strong>X_Seq-YYYYMMDDHHMMSS</strong>-snapshot”.<br> E.g.:<br> Background image: 12_13-20190704172200-snapshot.jpg<br> Empty label file: 12_13-20190704172200-snapshot.txt <em>Images annotated with insects are named:</em><br> “<strong>SZ_IP-MonthDate_C_Seq-YYYYMMDDHHMMSS</strong>”.<br> E.g.:<br> Image file: S1_146-Aug23_1_156-20190822133230.jpg<br> Label file: S1_146-Aug23_1_156-20190822133230.txt <strong>Abbreviations</strong>: <strong>YYYYMMDDHHMMSS </strong>– Capture timestamp with year, month, date, hour, minutes, and second<br> <strong>Seq</strong> – Sequence number created by the motion program to separate images<br> <strong>C</strong> – Identification of two cameras with Id=0 or Id=1 in system identified by <strong>SZ_IP</strong><br> <strong>MonthDate </strong>– Folder name for where the original image were stored in the system<br> <strong>SZ_IP</strong> – Identification of five camera systems: S1_123, S2_146, S3_194, S4_199, S5_187 (Two cameras in each system)<br> <strong>X</strong> – An index number related to a specific camera and folder ensuring unique file names of background images from different camera systems.<br> <br> The important information in a filename is system (<strong>SZ_IP</strong>), camera Id (<strong>C</strong>) and timestamp (<strong>YYYYMMDDHHMMSS</strong>). <strong>The three best YOLOv5 models (YOLOv5models.zip) from the paper are available in pytorch format.</strong> All models are tested with YOLOv5 release v7.0 (22-11-2022): ultralytics/yolov5: YOLOv5 in PyTorch <strong>insect1201-bestF1-640v5m.pt</strong>: Model no. 6 in Table 2 (F1=0.912)<br> <strong>insect1201-bestF1-1280v5m6.pt</strong>: Model no. 8 in Table 2 (F1=0.925)<br> <strong>insect1201-bestF1-1280v5m6.pt</strong>: Model no. 10 in Table 2 (F1=0.932) <strong>insects-1201val.yaml</strong>: YAML file with label names to train YOLOv5 <strong>trainInsects-1201m.sh</strong>: Linux bash shell script with parameters to train YOLOv5m6<br> <strong>valInsectsF1-1201.sh</strong>: Linux bash shell script with parameters to validated models