Deep Learning Framework

Keras and the R package ](keras)

TensorFlow, pyTorch and MXNet

General purpose object detection

(for an introduction about this section, see this page)

YOLO In YOLO a single convolutional network predicts the bounding boxes and the class probabilities for these boxes. Widely used, in particular with animals as "object" to detect. arXiv preprint

Tutorial by @vm-lbbe and @Gasp34 available here.

Mask-R-CNN The model generates bounding boxes and segmentation masks (a mask is the exact contours of an objects) of objects in the image. The mask detection requires a training on annotated images with the contours of all objects (animals for instance). arXiv preprint

Tutorial by @vm-lbbe and @Gasp34 available here.

RetinaNet New loss called Focal Loss to correct the problem of imbalance between foreground and background classes. arXiv preprint

Tutorial by @vm-lbbe and @Gasp34 available here.

Camera traps

Camera Trap Classifier This repository contains code and documentation to train and apply CNN for identifying animal species in photographs from camera traps. Code in Python and based on Tensorflow.

Tutorial by @vm-lbbe and @Gasp34 available here.

Microsoft codes including a too dedicated to camera traps with Megadetector One-class animal detector (i.e. detects any animal but does not classify) trained on several hundred thousand bounding boxes from a variety of ecosystems.

Tutorial by @vm-lbbe and @Gasp34 available here.

MLWIC2: Machine Learning for Wildlife Image Classification in R This package identifies animal species in camera trap images by implementing the model described in Tabak et al, MEE 2018 and updated in Tabak et al,biorxiv 2020 Linux, Mac, (Windows)

DLCTI Along with Norouzzadeh, PNAS 2018


Animal Scanner Software for classifying humans, animals, and empty frames in camera trap images. Along with Yousif et al, Eco Evo 2019


Face recognition

face_recognition Recognize and manipulate faces from Python.

Depth estimation


Training and testing depth estimation model

Object Tracking


Simple Online and Realtime Tracking with a Deep Association Metric

Unsupervised clustering

DeepCluster Clustering for Unsupervised Learning of Visual Features.

Model interpretation

tf-explain Interpretability methods as Tensorflow 2.0 callbacks to ease neural network's understanding.

Model interpretability with Integrated Gradients in Keras Integrated Gradients is a technique for attributing a classification model's prediction to its input features.

results matching ""

    No results matching ""