In some regards, veterinary medicine can be compared with human medicine in third world countries. Many basic laboratory tests are still too expensive for pet owners or are simply not feasible due to a lack of automated methods and the sheer number of different species. Deep learning can provide a solution to this and similar problems and bring much cheaper and more accurate diagnoses to veterinary medicine. In doing so, I believe that this technology can enable veterinary medicine to finally catch up with human medicine.
The use of deep learning to perform a wide variety of tasks related to medical imaging only requires suitable data to train readily available models, and this is why veterinarians need to be involved. Such problems can only be solved by people who are familiar with them. Computer scientists won’t be able to solve the majority of problems in veterinary medicine, but veterinarians can, not just because they are the ones who are confronted with these problems every day, but also because they have the medical expertise needed for collecting and labelling medical images.
This is a list of open source veterinary image datasets from which the whole veterinary profession and public could benefit. If you create any projects using the datasets below, please get in touch with me at k.vinicki@gmail.com as I would love to hear about it.
| Name | Description |
|---|---|
| Canine Coccidiosis | This dataset contains images and labels of Isospora canis and Isospora sp. oocysts, a coccidian parasites that infect intestinal tract in dogs |
| DogTeethAge | This dataset contains 44 dog teeth images together with their ages |
| Feline reticulocytes | This dataset contains 1086 images and labels of feline reticulocytes. This data was used as the basis of the following paper - Using Convolutional Neural Networks for Determining Reticulocyte Percentage in Cats |
| Ixodidae ticks | This dataset contains images of ticks from six genera |