search

janerikhy / geoDjango

0 stars 0 forks source link

Implement Object Detection and Recognition #1

Open janerikhy opened 3 years ago

janerikhy commented 3 years ago

What?

Currently, there is only implemented a jS script that tells what the image contains on loading. It is desirable to have object detection and categorization to work as the user uploads an image.

How?

Make use of the Google Vision API to train a model. The model will be imported and used in observations/utils.py. Must probably alter the ObservationTest class, as it of now does not contain any species column in the database.

janerikhy commented 3 years ago

The ML-model will be created with TensorFlow due to payment requirements from Google Vision API. However, if decided to go back to Google API, Vertex API should be used.

janerikhy commented 3 years ago

With the limited timespan of this project, it is probably best to only develop the classification algorithm and let the object detection part be a topic of further work.