Hello Everyone,
I’m now working on a project for wildlife conservation that tracks and monitors numbers of animals in a protected region using computer vision.
Identifying and locating different wildlife species that are filmed by motion-activated cameras positioned across the reserve is the main objective.
Below is a quick synopsis of my venture and the difficulties I’m currently facing:
Dataset: The distribution of the about 10,000 labelled photos I have of various animal species is rather uneven. For example, compared to uncommon animals like foxes and badgers, there are many more photographs of deer and birds.
Model: I’ve already completed multiple training and validation cycles with a pre-trained YOLOv5 model. The model does not work well on less common animals, but it does function rather well on common species.
Problems:
Class Imbalance: The model is biased towards fewer common species as a result of the unbalanced dataset.
Negatives/False Positives: Occasionally, especially for rare species, the model misidentifies an animal or fails to detect it at all.
Environmental Variability: The identification procedure becomes more challenging when images are captured at various moments of the day, in different kinds of weather, and with variable backdrops.
Inquiries for the Community:
Keeping the Dataset Balanced: How can I address a category discrepancy in my dataset? 
Is artificial intelligence data augmentation useful, and if so, how can I go about utilising it?
Enhancing Rare Species Assessment: Are there any particular tactics or model modifications that could aid in raising the accuracy of rare species detection?
Managing Environmental Variability: How should a model be trained to be as broadly applicable as possible in a variety of environmental circumstances?
I also checked this 
https://www.nature.com/articles/s41467-022-27980-uipath
Your opinions, recommendations, and any resources you might offer would be very valued. If you’ve had any comparable situations, please share them with me, along with how you handled them.
I appreciate 
your assistance in advance!