The concept: The idea is to combine visual plant identification technology with a large database of ethnobotanical (plant utilization) information so that the user could instantly know which plants from their natural surroundings can be used and how along with the information when/how to collect and prepare them for use. Two main use categories are edible and medicinal plants, but all the other possible uses would be included (e.g. technical, handicraft, veterinary, etc.). The app would be intended for use in the wild or semi-wild plant habitats (not for garden plants).
The novelty of it: The novelty of the idea lies in the app being able to identify many plant species quickly from a single image or footage (with the help of supporting information like geolocation) rather than only specific species from a specific photo. And also in vastly enlarging the ethnobotanical database available to the app so that the majority of wild plants in any area would have some uses attributed to them. The app would therefore be able to quickly present you with the list of plant species present around you and information on how each species can be used. Practical guides, such as videos on plant collection and preparation would also be included.
Visual identification part: Using device's camera the person would "scan" a certain place in the wilderness. The most convenient approach would be to simply film or make an extended panoramic photo around yourself. There are already apps for plant identification from photos and they are pretty accurate. With the improved image recognition technology, it would be possible to identify the type of the habitat as well as separate plants from a high enough quality panoramic photo or footage.
Identification of the habitat type (which is relatively easy to do from the views of the surroundings) in combination with the information of the geographic location of the person would already let the app guess what species of plants can be present in the surroundings and make it easier for the algorithm to then recognize separate plants by analyzing specific parts of the image or footage.
The app could then also guide the user to approach more specific areas and take additional photos from closer distances to further drive the process of identifying all the species present in the surrounding habitat. If a person gets interested in a particular plant, they could then take additional photos up-close to help the algorithm identify the species with higher certainty.
Wild plant habitat mapping is a common practice in environmental science and the app could also use already available information (databases of habitat mappings) to supplement the identification process. Plant species are usually repeating over a relatively large land area (which makes up a habitat) so when the species in a certain patch of the habitat are identified the person then can walk around and they will most likely find the same species.
A simple smartphone camera might not be enough for accurate large-scale identification, so a separate, specialized camera (like a 360° camera on a stick) paired with a smartphone or other personal device might be necessary. In a more distant future scenario, the technology could also use material samples to identify the species using methods of genetics and molecular biology.
Utilization part: Some present plant identification apps already provide ethnobotanical information about some of the plant species, but still to a very limited extent. The envisioned app would use the information available in plant-use databases like pfaf, but additionally, many other literature sources (books, scientific articles) to compile a large enough information pool which then would be available to the user.
It's important to prioritize the information which is trustworthy and based on practical knowledge, ideally, it should be backed with either traditional knowledge (ethnobotanical research) or otherwise practically tested (books by plant enthusiasts, etc.). Specific use being mentioned in multiple literature sources is also a criterion for trustworthiness. Reliability rank of an information about specific use of a plant should be shown to the app user in "plant's profile".
To make plant utilization as easy as possible, information about the practical aspect (how and when to collect the plant or its parts and how to prepare and consume or otherwise utilize it) should be presented in a most learning-efficient and user-friendly manner. Therefore extensive use of educational videos seems like the best approach.
The danger of poisonous plants:The plants that are seriously poisonous are not that common. But they are the ones that should be given exceptional emphasis in the identification algorithm. If a person decides to try using a certain plant they should then additionally take several photos of that plant up-close for the algorithm to completely rule out the similarities with poisonous plants and ensure safety at the same time identifying the species with higher accuracy.
Endangered plants: This kind of technology could make things easier for people who are willing to abuse wild resources. Some possible solutions to this: pop-ups and emphasized information near descriptions of certain plants about their endangered status and warnings about fines associated with their abuse (for starters) and app use only allowed through logging in with personal ID, if things would get out of control. Although it's not likely that a lot of people would use the app extensively, some, I imagine, would use it for playing and others (like nature enthusiasts, hikers, low-budget travelers, or people who live in rural settings and more or less depend on wild nature for survival) would use it more extensively.
Why is it important:
Responsible use of wild resources reduces the consumption of farmed food as well as other manufactured products, which in turn lowers capitalistic consumerism, increases the possibility of replacing agricultural lands with natural or semi-natural ecosystems which then reduces CO2 emission, improves air quality, and as a feedback loop provides more natural products to depend on.
Making utilization of wild plants easier for laypeople increases public interest and appreciation of nature, for they get to know how it can be useful practically and that the possibilities of utilization are surprisingly abundant. All it takes is easily accessible information to simplify the process enough.
Responsibly used natural resources provide a possibility for free off-grid living and enable individuals to be independent and self-reliant.
I have a friend who has been working on the visual identification part for the company he owns
It's obviously not used for the same purpose as you are proposing and it focuses on fruit. They use it to get an estimated yield and to identify problem areas but I believe the technology might be similar to what you will need.
I know you mentioned that similar technology already exists, but it might be worth taking a look at what they're working on. Here's a video of their AI tech at work.
If this can be linked to a database of the information you mentioned your app will be good to go!