Data

Tools

Indexes

Applications

Experiments

Open Source Science

Recent Releases of insect-detect

insect-detect - insect-detect-v1.7

Camera trap paper now published in DOI PLOS ONE

The v1.7 release includes several new features and updates:

New features

  • Add option to save full HQ frames at specified frequency, independent of detections (e.g. for training data collection) in https://github.com/maxsitt/insect-detect/commit/1ef2ded71a28acec435015cc30b313ce061cdd26
  • Add support for Witty Pi 4 L3V7 as alternative power management board to the PiJuice Zero in https://github.com/maxsitt/insect-detect/commit/230c14b398b44e3075ae23a23e815d982a072ffb
  • Add utils package with modules providing utility functions to improve maintainability in https://github.com/maxsitt/insect-detect/commit/8217d5d47af27ec94a53ca4a0839d26de92d5b23
  • Add option to set auto focus range (restrict auto focus to a minimum and maximum distance measured from camera) in https://github.com/maxsitt/insect-detect/commit/d5ecf04c309b5d63977f47c2b42d06a720c3e25d
  • Add option to store all captured data in an uncompressed .zip file in https://github.com/maxsitt/insect-detect/commit/5737451d11ec7dd82644812baa33327011f4f62e
  • Add option to use bounding box coordinates from detections to set auto exposure region in https://github.com/maxsitt/insect-detect/commit/7d4eaf1bcdf8c151ee7c05269e06d81229333fad
  • Add installation instructions and example scripts to use the DepthAI SDK in https://github.com/maxsitt/insect-detect/commit/6968c1e9f0726b4c14d59a61c24745d77ee8f0b4
  • Add option to save cropped detections with aspect ratio 1:1 (default) or with original variable aspect ratio in https://github.com/maxsitt/insect-detect/commit/e803d7e4cc8b1a88911facb892e840758311052a and https://github.com/maxsitt/insect-detect/commit/1716baaa5da6261c0dbd89b54328a5452472286a

Important Updates

  • Handle external shutdown trigger (e.g. pressing of button connected to Witty Pi while script is running) in https://github.com/maxsitt/insect-detect/commit/2690956dd54b544a78f00b5cdeb6e40a14520e36
  • Increase pipeline speed if full frames are saved additionally by using threading in https://github.com/maxsitt/insect-detect/commit/1ef2ded71a28acec435015cc30b313ce061cdd26
  • Optimize recording ID creation in https://github.com/maxsitt/insect-detect/commit/7d970d692061f70ce9b3477d776a1f2bdc1c0dc3
  • Use ISO 8601 format for timestamps in https://github.com/maxsitt/insect-detect/commit/8217d5d47af27ec94a53ca4a0839d26de92d5b23
  • Use new Sync node instead of Script node to synchronize HQ frames with tracker output in https://github.com/maxsitt/insect-detect/commit/e3101c0e982bc854be4b77dce767e80ef0b18657

Changelog

Changes between previous release and this release: https://github.com/maxsitt/insect-detect/compare/v1.6...v1.7

- Python
Published by maxsitt almost 2 years ago

insect-detect - insect-detect-v1.6

In the v1.6 release, new YOLOv6n, YOLOv7-tiny and YOLOv8n detection models were added to the existing YOLOv5n model.

| Model | size
(pixels) | mAPval
50-95 | mAPval
50 | Precisionval
| Recallval
| SpeedOAK
(fps) | params
(M) | | ----------- | --------------------- | -------------------- | ----------------- | --------------------- | ------------------ | ---------------------- | ------------------ | | YOLOv5n | 320 | 53.8 | 96.9 | 95.5 | 96.1 | 49 | 1.76 | | YOLOv6n | 320 | 50.3 | 95.1 | 96.9 | 89.8 | 60 | 4.63 | | YOLOv7-tiny | 320 | 53.2 | 95.7 | 94.7 | 94.2 | 52 | 6.01 | | YOLOv8n | 320 | 55.4 | 94.4 | 92.2 | 89.9 | 39 | 3.01 |

These models can reach up to 60 fps inference speed (YOLOv6n) on the OAK-1 device with 1080p sensor resolution.

New features

Add new YOLO models: https://github.com/maxsitt/insect-detect/commit/46ede0891eb85850699fe8c79b4e2ec0bc03a6f2

Important Updates

  • Update yolo_preview.py and yolo_tracker_preview.py to use 1080p as default sensor resolution, which leads to higher fps of the detection models without decreasing quality of the downscaled preview frames: https://github.com/maxsitt/insect-detect/commit/135e75f8f4d0b2e6e5764f1c2c5f2eecf052fd4f and https://github.com/maxsitt/insect-detect/commit/066587c548339ba1d40c2b46e4ee6aa4bc407d9a
  • Update yolo_tracker_save_hqsync.py and yolo_tracker_save_hqsync_nopj.py to use 1080p as default resolution for the synced HQ frames, which increases inference and pipeline speed to ~12 fps which is fast enough to track moving insects. Save overlay only once with all detected insects drawn on the frame: https://github.com/maxsitt/insect-detect/commit/ac1162e420e46043ca8b9777fee3b67e6688a733 and https://github.com/maxsitt/insect-detect/commit/5eea522c4e6ef6484d9c16e17654385687149b16
  • Update video_capture.py to use ISP scaling for 1080p resolution, which leads to higher quality frames compared to setting the sensor resolution to 1080p: https://github.com/maxsitt/insect-detect/commit/e06624ba7431e24682f35399a14677eca1e33510

Changelog

Changes between previous release and this release: https://github.com/maxsitt/insect-detect/compare/v1.5...v1.6

- Python
Published by maxsitt almost 3 years ago

insect-detect - insect-detect-v1.5

For the v1.5 release, a new YOLOv5n detection model, trained on downscaled 320x320 frames was added. With the new model and the adapted scripts (320x320 model input), performance increases significantly to 40 fps for insect detection + tracking on the OAK device, without affecting accuracy. The higher framerate now makes ZERO_TERM_IMAGELESS the object tracker type with the best performance, which increases tracking accuracy for fast moving insects significantly.

| Model
(.blob) | size
(pixels) | mAPval
50-95 | mAPval
50 | Precisionval
| Recallval
| Speed
OAK
(fps) | | ----------------------- | --------------------- | -------------------- | ----------------- | --------------------- | ------------------ | -------------------------- | | YOLOv5n (+ tracker) | 320 | 53.9 | 97.6 | 96.0 | 96.6 | 40 | | YOLOv5n (+ tracker) | 416 | 58.2 | 97.4 | 97.0 | 95.0 | 30 | | YOLOv5s (+ tracker) | 416 | 63.4 | 97.8 | 96.6 | 95.6 | 17 |

New features

Add new YOLOv5n model: https://github.com/maxsitt/insect-detect/commit/0091a58687b87b46081825db64b0733e1f61cba2

Important Updates

  • Update yolov5_preview.py: https://github.com/maxsitt/insect-detect/commit/0a40d3bebfe1879eb09bad328c335e9857804649
  • Update yolov5trackerpreview.py: https://github.com/maxsitt/insect-detect/commit/140db67eae4ce3de22632215d185844c61f697cd
  • Update monitoring scripts: https://github.com/maxsitt/insect-detect/commit/2f3e7382f7807fddfc6d109a16cb0cd419f828a1

Changelog

Changes between previous release and this release: https://github.com/maxsitt/insect-detect/compare/v1.4...v1.5

- Python
Published by maxsitt about 3 years ago

insect-detect - insect-detect-v1.4

For the v1.4 release, new YOLOv5n detection models were added. With the YOLOv5n models performance increases significantly to 32 fps for insect detection and 30 fps for detection + tracking on the OAK devices. To reproduce the correct speed (fps) measurement while connected via SSH (X11 forwarding of the frames), print fps to the console and comment out cv2.imshow(), as this will significantly slow down the received message output and thereby fps. If you are using e.g. a Raspberry Pi 4 B connected to a screen, fps will be correctly shown in the livestream.

| Model
(.blob) | size
(pixels) | mAPval
50-95 | mAPval
50 | Precisionval
| Recallval
| Speed
OAK
(fps) | | ---------------------------- | --------------------- | -------------------- | ----------------- | --------------------- | ------------------ | -------------------------- | | YOLOv5n (5 shaves) | 416 | 58.2 | 97.4 | 97.0 | 95.0 | ~32 | | YOLOv5n (4 shaves) + tracker | 416 | 58.2 | 97.4 | 97.0 | 95.0 | ~30 | | YOLOv5s (5 shaves) | 416 | 63.4 | 97.8 | 96.6 | 95.6 | ~17 | | YOLOv5s (4 shaves) + tracker | 416 | 63.4 | 97.8 | 96.6 | 95.6 | ~17 |

New features

Add new YOLOv5n models: https://github.com/maxsitt/insect-detect/commit/5fb3e4854aeb343caab5b353a4fbc105f0214c77

Important Updates

  • Update YOLOv5 preview scripts: https://github.com/maxsitt/insect-detect/commit/fd943bcb761b3198eea3ebecb5cf29498b91140b
  • Update monitoring scripts: https://github.com/maxsitt/insect-detect/commit/01f31489435faeba112b9fc68e9f24fe71ece340 and https://github.com/maxsitt/insect-detect/commit/5d3714a754fea327f676c3f1999c65c10330da6e

Changelog

Changes between previous release and this release: https://github.com/maxsitt/insect-detect/compare/v1.3...v1.4

- Python
Published by maxsitt about 3 years ago

insect-detect - insect-detect-v1.3

In this release the video_capture.py script to save encoded HQ frames (1080p or 4K resolution) with HEVC/H.265 compression to .mp4 video file was added. As there is no encoding happening on the host (Raspberry Pi), CPU/RAM usage is minimal. This makes it possible to record 30 fps 4K video without any load on the Raspberry Pi Zero 2 W.

New features

Add video_capture.py script: https://github.com/maxsitt/insect-detect/commit/7f53f006c3bc36f64452c1de480b1adb1b7a5027

Important Updates

  • Update monitoring scripts: https://github.com/maxsitt/insect-detect/commit/e1d7745b37b677809da0c452499381e913467166
  • Update frame capture scripts: https://github.com/maxsitt/insect-detect/commit/4e46f2f918f73753aa604ff16a89a90df1ea9716
  • Update still_capture.py: https://github.com/maxsitt/insect-detect/commit/3254d88e237e664b99b0ebd603a89e6f869f6128

Changelog

Changes between previous release and this release: https://github.com/maxsitt/insect-detect/compare/v1.2...v1.3

- Python
Published by maxsitt about 3 years ago

insect-detect - insect-detect-v1.2

In this release the still_capture.py script to save still images in the highest possible sensor resolution was refactored, now you can set the recording time with the optional argument -min (e.g. -min 5 for 5 minutes recording time).

The frame_capture.py script was added, to save HQ frames (e.g. 3840x2160 px), optionally together with LQ frames (e.g. 416x416 px) to .jpg at specified time interval. The recording time can be set with the optional argument -min.

Important Updates

Refactor still_capture.py script: https://github.com/maxsitt/insect-detect/commit/3864e5f012b74ce8b561bf08765db4c2de961cc5

New features

Add frame_capture.py script: https://github.com/maxsitt/insect-detect/commit/c7cfeea8054886860feaa699d6243a9e97de0bee

Changelog

Changes between previous release and this release: https://github.com/maxsitt/insect-detect/compare/v1.1...v1.2

- Python
Published by maxsitt about 3 years ago

insect-detect - insect-detect-v1.1

Release for archiving in Zenodo.

- Python
Published by maxsitt about 3 years ago

insect-detect - insect-detect-v1.0

Initial release of Python scripts and YOLOv5s detection model (.blob format) for automated insect monitoring with the Insect Detect DIY camera trap.

- Python
Published by maxsitt about 3 years ago