Home

Overview

This project originates from a thread on reddit

The goal is to create an application which would be able to identify and detect individual fallow deer in a video. To be more precise, each individual fallow deer have their own distinct fur pattern and as far as I can tell from real life observations, they keep it throughout their entire life. So we’re not talking about a classification “this is a deer”, but rather - like human face recognition is mapping to a person: this is “Distinct Doe 1”, “Distinct Doe 2” and so on. The goal is to recognize if this particular deer has already been present in earlier footage and if not, save it for later reference.

Caveat 1: patterns on the left and right side of the animal are not equal. Also, once they get their winter fur the patterns become invisible, but they do reemerge when they change back to the summer fur.

Caveat 2: as the animal grows, the density of the pattern decreases, i.e. the dots and lines look "compressed" on a fawn, but move further away from each other as the deer get older. The pattern as such remains , i.e. the orientation of lines, number of dots and their relative positions to each stays the same, but the canvas is being stretched so to speak.

The processed results need to be saved in a database alongside various metadata: when has this deer appeared the first time, when it was seen for the last time, and so on, which would allow to count and track the population which gets caught on camera.

Possible Recognition Strategies

Currently two strategies are being discussed:

• Identify the animals by face recognition, this approach is taken by the Bear ID Project This approach would eliminate the problematic of different patterns on each side of the animal and also the seasonal change in the fur, where the pattern becomes invisible in the winter. It is however unclear if deer provide enough features to identify them solely by their faces.
• Identify the animals by their fur patterns, this is seems somewhat similar to the Whale Recognition Project

Useful Links

• https://bearresearch.org/ face recognition for bears

  • https://onlinelibrary.wiley.com/doi/10.1002/ece3.6840
  • https://github.com/hypraptive/bearid

• https://erdem.pl/2021/02/animal-recognition-with-siamese-networks-and-mean-embeddings - recognition of farm animals

  • https://www.frontiersin.org/articles/10.3389/fnbeh.2022.759943/full

• https://www.kaggle.com/c/noaa-right-whale-recognition recognizing whales by their unique patterns

  • https://www.kaggle.com/competitions/happy-whale-and-dolphin

• https://dinov2.metademolab.com/ Learning Robust Visual Features without Supervision

  • https://arxiv.org/pdf/2304.07193.pdf

• https://universe.roboflow.com/george-6lzhu/deer-xcinb/dataset/1 object location model

• https://www.zhihu.com/column/FaceRec references on face recognition problems (in Chinese)

• https://www.sciencedirect.com/science/article/pii/S2405896319324772 Individual Cattle Identification Using a Deep Learning Based Framework

• https://ieeexplore.ieee.org/abstract/document/7532404/ Automatic individual holstein friesian cattle identification via selective local coat pattern matching in RGB-D imagery

• https://www.sciencedirect.com/science/article/pii/S1574954123003631 paper on deer face recognition

• https://arxiv.org/pdf/2005.02905.pdf paper on individual animal pattern recognition (focuses on zebras, jaguars and alike)