mobilenet-wildlife-detector by viam-labs - Registry

mobilenet-wildlife-detector

by viam-labs

Public

A neural network model to detect animals from video or images. It is similar to the megadetector. However it is based on SSD-MobilnetV2 instead of Faster-RCNN + InceptionResNetv2. This way, it can be used with TFLite since Faster-RCNN is not supported as of June 2020 in TFLite. Of course the accuracy will not be as good as the megadetector, but the inference speed is much faster. It was trained using the ENA-24 dataset and the Caltech Camera Traps dataset using transfer learning (only the last few layers of the Mobilenet is re-trained). Note that it model will not classify the type of the animal. GitHub Link

Model framework

TFLite

Task type

Object detection

Object detection: Predict object labels and their location within an image.

Files

• detect.tflite
• labelmap (1).txt

Size

4.52 MB

Latest version

2024-02-01T15-03-42•02/01/2024

Resource type

ML model

Usage

1
org • 1
machine