https://github.com/elucidation/bird_viewer

View bird events from camera
https://github.com/elucidation/bird_viewer

Last synced: about 1 year ago
JSON representation

View bird events from camera

• Host: GitHub
• URL: https://github.com/elucidation/bird_viewer
• Owner: Elucidation
• License: mit
• Created: 2024-01-02T18:14:37.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2024-01-07T01:04:01.000Z (over 2 years ago)
• Last Synced: 2025-02-10T00:16:39.772Z (over 1 year ago)
• Language: Jupyter Notebook
• Homepage: https://www.tetralark.com/bird_viewer/
• Size: 13 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Birb overview

I have a camera that tracks and captures small crop images of motion objects including birds.

It sees several types of birds (crows, juncos, finches, sparrows etc.) which I'd like to classify.

![Predictions](predictions.png)

# Classfication Model Training

[Model training ipython notebook](bird_classify.ipynb)

Trained on the [Caltech-UCSD Birds-200-2011 (CUB-200-2011)](https://paperswithcode.com/dataset/cub-200-2011) dataset. It contains 11,788 images of 200 subcategories belonging to birds, 5,994 for training and 5,794 for testing.

We take a MobileNetV2 architecture with the last layer removed, replaced with dropout -> dense(200) -> dropout -> dense(70) for 70 groups of birds.

![model](model.png)

Generally it performs with 80% accuracy on the test dataset.

# Camera events

```json

[

    {"area":null,

    "box":null,

    "camera":"birbcam",

    "end_time":1704213300.999055,

    "false_positive":null,

    "has_clip":true,

    "has_snapshot":true,

    "id":"1704213243.00595-x47jwu",

    "label":"bird",

    "plus_id":null,

    "ratio":null,

    "region":null,

    "retain_indefinitely":false,

    "start_time":1704213238.00595,

    "sub_label":null,

    "top_score":0.82421875,

    "zones":[],

    "thumbnail":""},

    // ... more entries 

]

```

The camera provides the cropped images and associated metadata as an events JSON data file.

First, lets make a webpage that can load this JSON events file and do some simple visualizations.

![Example page](example_page.jpg)

## Running

`python -m http.server`