InsectAI Model Zoo
A registry of models for insect detection, classification and trait quantification. Models must be vision-based, must have a scope spanning multiple insect genera, and must have open or at least lightly gated weights. Weights stay with their authors, this catalogue points at them and never redistributes them.
⚠️ Displayed licenses are copied from the original authors' work and may not show the full picture. They may differ or be outdated from the actual terms of use, which you must independently verify on each model's official page.
- 29 models
- 4 purposes
- 9 tasks
| Model | Purpose | Task | Architecture | Input | Vocabulary | Input size | Year | License | Output Format |
|---|---|---|---|---|---|---|---|---|---|
| AMI insect detector | Insect detection | Object Detection | Faster R-CNN ResNet50-FPN, Faster R-CNN MobileNetV3-Large-FPN | image | closed | any | 2023 | AGPL-3.0-only | ami-detector-boxes |
| AMI moth classifiers | Insect classification | Binary Classification, Classification | ResNet50, ConvNeXt-T | image | closed | 128x128 (Mila models), 300x300 (Turing models) | 2024 | AGPL-3.0-only | — |
| ArthroNat | Insect detection | Object Detection | YOLO11 | image | open | 640x640 | 2026 | AGPLv3 | ultralytics-detect-results |
| BEEomass | Trait quantification | Regression | EfficientNetV2 | image | open | 224x224 | 2026 | CC-BY-4.0 | — |
| BeetleFlow | Trait quantification | Semantic Segmentation | Mask2Former, Swin Transformer (Swin-L) | image | closed | 512x512 | 2025 | MIT | beetleflow-color-mask |
| BioCAP | Insect classification, Model embeddings | Classification, Embedding | OpenAI CLIP ViT-B/16 | image, text | open | 224x224 | 2025 | MIT | — |
| BioCLIP | Insect classification, Model embeddings | Classification, Embedding | ViT-B/16 (CLIP) | image, text | open | 224x224 | 2023 | MIT | — |
| BioCLIP 2 | Insect classification, Model embeddings | Classification, Embedding | ViT-L/14 (CLIP) | image, text | open | 224x224 | 2025 | MIT | — |
| BioMoth | Insect detection, Trait quantification | Object Detection, Counting, Regression | YOLO11s | image | closed | 640x640 | 2026 | Apache-2.0 | biomoth-csv |
| BioTrove-CLIP | Insect classification, Model embeddings | Classification, Embedding | CLIP | image, text | open | 224x224 | 2024 | MIT | — |
| Bjerge et al. (2023) | Insect detection | Object Detection, Classification | YOLOv5 | image | closed | 1280x1280 | 2023 | CC-BY-4.0 | — |
| Bjerge et al. (2025) | Insect detection, Insect classification | Object Detection, Classification | YOLOv5, ResNet50 | image | closed | 1280x1280 | 2025 | CC-BY-4.0 | mcc24-csv |
| CLIBD | Insect classification, Model embeddings | Classification, Embedding | ViT-B/16, BarcodeBERT, BERT-Small | image, text | open | 224x224 | 2025 | MIT | — |
| ConvNeXt (base) | Model embeddings | Classification, Embedding | ConvNeXt-Base | image | closed | 224x224 | 2022 | Apache-2.0 | — |
| DINO v3 | Model embeddings | Embedding | Vision Transformer (ViT), ConvNeXt | image | closed | any | 2025 | custom | — |
| flatbug | Insect detection | Object Detection, Instance Segmentation | YOLOv8-seg, YOLO26-seg | image | closed | any | 2026 | MIT | flatbug |
| Grounding DINO | Insect detection, Model embeddings | Object Detection, Classification, Embedding | Swin Transformer | text, image | open | any | 2023 | Apache-2.0 | grounding-dino-hf-results |
| Insect Detect platform | Insect detection | Object Detection, Classification | YOLOv6 | image | closed | 800x448 | 2026 | GPL-3.0-or-later | insect-detect-csv |
| Insect-Foundation (vision backbone) | Model embeddings | Embedding | ViT-B/16 | image | closed | 224x224 | 2024 | Not found | — |
| InsectDCT | Insect detection, Insect classification | Object Detection, Classification, Tracking | YOLOv11 | image | closed | 1920x1080 | 2026 | GPL-3.0-or-later | insectdct-csv |
| InsectMorphoAI | Trait quantification | Object Detection, Instance Segmentation, Regression | YOLOv8-obb, YOLOv8-seg | image | closed | any | 2026 | MIT | insectmorphoai-csv |
| lepinet | Insect classification | Classification, Embedding | ViT-L/14 (BioCLIP-2), ConvNeXt-L (DINOv3), EfficientNetV2-S | image | closed | 256x256 (ViT, resampled to 224 inside the model), 320x320 (ConvNeXt), 256x256 (EfficientNet) | 2026 | CC-BY-NC-4.0 | — |
| Mothbot Detect (MBD) | Insect detection | Object Detection | YOLO26-OBB | image | closed | 1600x1600 | 2026 | AGPL-3.0-only | mothbot-detection-json |
| OpenCLIP (ViT-B/16, LAION-2B) | Model embeddings | Classification, Embedding | CLIP with ViT-B/16 image encoder | image, text | open | 224x224 | 2023 | MIT | — |
| POLLINATOR — Serra-Marin et al. (2025) | Insect detection | Object Detection | YOLOv5m | image | closed | 1024x1024 | 2025 | AGPL-3.0-or-later | pollinator-frame-folders |
| SAM 3 | Insect detection, Model embeddings | Object Detection, Instance Segmentation, Tracking | ViT (ViTDet) + DETR detector + SAM 2 tracker | text, image, box, point, mask | open | 1008x1008 | 2025 | custom | — |
| Stark et al. (2023) | Insect detection, Insect classification | Object Detection, Classification | YOLOv5, YOLOv7 | image | closed | 640x640 | 2023 | AGPL-3.0-only | yolov5-detect-txt yolov7-detect-txt |
| TaxaBind | Insect classification, Model embeddings | Classification, Embedding | CLIP, with ViT-B/16 image encoder, plus other feature extractors. | image, text | open | 224x224 | 2025 | MIT | — |
| ultra-lightweight CNNs | Insect detection | Classification | MobileNetv2 | image | closed | 120×160 | 2025 | GNU GPLv3 | ecto-trigger-tflite-score |