Vision: A world that we can have image recognition more readily available to identify insects, animals, birds, reptiles, and various species with limited image data pools to train ML models.
Feature Request: There is an option to " Render (Local) to create a single image render from your different cameras", but this is slow, tedious, and does not work well in generating thousands of images to train ML models. Is Unreal Engine 5 good enough to take a virtual beetle that was either scanned or recreated from a photo into the unreal engine, then have unreal engine output X number of images of different angles, lighting, backgrounds, and depending on the model (let’s say a bird, can show the bird in flight, wings out, one wing up, different physical motions of the bird in conjunction with different angles, lighting, and backgrounds), so that we can train ML models to identify these species with a picture or in real time videos. This is huge especially when images for a particular species is not readily available or existent, and even more so, not practical at all to get the bird and place the bird in enough real world environments to train a ML model.
Long version: Unreal Engine 5 is perfect for this job. The images within Unreal Engine 5 are so realistic, that I believe they can generate massive amounts of photos to train ML models that can then identify real world objects based on thousands of “fake” images. If this becomes possible, then we can being to have things like training huge databases to identify and educate us about the world around us from individuals taking pictures, to students in school, students on a field trip, people in areas identifying poisonous or venomous species, to photographing and identifying rare and endangered animals, or for poor countries and individuals to identifying various sources of food like which frogs can be eating to reduce malnutrition in children for lack of protein and nutrients.