Yeah Na
Yeah an
Response from developer
Thanks for your feedback. We’re sorry about the incorrect ID issues. If you still have the photo(s), please send them to weeds@invasives.com.au to help investigate. We are working on an update which significantly increases the number of training images per species, so identification suggestions should be more accurate soon. This includes the training sets for Myriophyllum aquaticum and Erica arborea.
Clunky and misleading
Response from developer
Thanks for your feedback. We’re sorry the suggestions were inaccurate. We’re working on a new update which includes many changes to improve the ID component. Primarily substantially larger training sets for the model to learn on, nearly double the species coverage, and implementing a threshold of confidence, below which a record cannot be created due to uncertainty. We’re looking into ways we can make the species list easier to navigate (for disagreeing with the suggestion) including filtering options.
ID needs work
Response from developer
Thanks for your feedback. We’re working on a new update which includes many changes to improve the ID component. Primarily substantially larger training sets for the model to learn on, nearly double the species coverage, and implementing a threshold of confidence, below which a record cannot be created due to uncertainty.
Amazing
Response from developer
Hi there, thanks so much for the positive feedback! We will continue to work on ironing out inaccuracies, and keep an eye out for a new update soon with many more species!
Absolutely unreliable
Response from developer
Hi there, thanks for your feedback. We’re sorry that WeedScan did not identify weeds well for you. WeedScan uses a live updating image recognition engine that provides suggestions, and the speed with which these suggestions update can be tailored using the settings menu (under “Weed Update Frequency”). If it is cycling through suggestions quickly then it’s likely that it is quite unsure of species, and thus a less reliable estimate. This is usually because the species is not covered by the WeedScan AI model, or the image is being captured from too far away from the plant. Regarding accuracy, if you have specific examples, we could review please get in touch with us at weeds@invasives.com.au, it would help us identify where to improve for future updates.
Its always wrong
Response from developer
Hi there, thanks for your feedback. We’re sorry that WeedScan did not identify weeds well for you. WeedScan is currently trained on 459 priority weeds and lookalike species which are viewable on our species list. We have found from reviewing the observations that many users are photographing plants that are not on this list, predominantly common lawn and garden weeds. We are working to address this issue, but if you have specific examples, we could review please get in touch with us at weeds@invasives.com.au.
Great idea still needs work
Response from developer
Hi there, thanks for the feedback! We’re sorry that WeedScan did not meet your expectations. Regarding the plants wrongly identified as Lantana, did you know which plants they were? There are many common garden plants that are not covered by the WeedScan model with quilted looking leaves which closely resemble Lantana. If the app has suggested Lantana but with low confidence then it is likely that the plant isn’t Lantana but a species that isn’t in the AI model. Apologies for a slow response to your bug reports at the time, the staff member who manages the mailbox was in the field for a few weeks, but I believe you should have received a response afterwards. If this is not the case, please get in touch. If you keep experiencing bugs with the app or the image recognition model, please get in touch with us at weeds@invasives.com.au (mentioning your device type if possible).
Pretty good but some rough edges
Cons: It is obviously a version 1.0 product and needs a few fixes. In particular, the experience of anonymous weed reports could be better as there is not much feedback that the report has landed in the appropriate place
Response from developer
Hi there, thanks for the feedback and we’re glad you’re liking WeedScan so far. Yes, WeedScan is trained on leaves as well as flowers and fruit, and only able to identify the 459 species that it knows. We appreciate and will consider your feedback on anonymous weed reports for future updates.
Useless
Response from developer
We are sorry that you had a negative experience using WeedScan. WeedScan was primarily designed to identify priority weeds in environmental and agricultural contexts and is not intended to identify many common garden weeds. You may want to have a look at our list of identifiable weeds to make sure the weeds in your garden are also in our AI model.