automated picture recognition on the Danish Fungal Atlas
Screenshot from the web version of the automated picture recognition on the Danish Fungal Atlas website. My uploaded Jelly Ear suggests a range of options, the first one beign correct.

New mushroom identification by photo recognition web site and app – impressive results!

There have been several mushroom identification apps / web sites around for a while. Some are:

  • Digital versions of an established key – for example the key in Roger Phillips’ Mushrooms book moved to a web site and an app with photos and descriptions. Roger’s app was around for many years but has disappeared.
  • Visual keys. The MycoKey MMI ® (Morphing Mushroom Identifier) is a great example. It is described as an innovative identification tool which models your fungus on screen as you enter the characters with automatic presentation of the most likely species. A more detailed version of MycoKey to install on your PC is also sold.
  • This category of mushroom identification apps perform identification based on a picture you take or upload. There have been apps before that have claimed to identify them from a photo alone, but experts have dismissed some as “potentially deadly“. You know that one random example of a particular mushroom species you encounter can differ from that wonderful specimen in a book . How young or old the specimen is, what the weather has been doing – washing out colours etc. all means this is a big ask.

A new web site / app takes mushroom photo recognition to a new level. In this blog post we try it out. Overall, I am very impressed. Yes, it needs to be used with caution and a good level of knowledge and use of several books is still wise. Might we make those numerous forum posts / emails – “I picked this, can I eat it?” a thing of the past? For more on mushroom identification look at this previous blog post of ours.

New Danish Website and App

This Danish website’s new tool and related app uses automated picture recognition and artificial intelligence (AI). It was trained on images from the excellent Danish Fungal Atlas. The system has been developed by Milan Šulc og Professor Jiri Matas from Czech Technical University in Prague (CTU), Lukáš Picek from University of West Bohemia (UWB) and the Danish Fungal Atlas. The latter group includes the respected Danish fungi authors of an amazing set of fungi books I am contemplating (Fungi of Temperate Europe (2-Volume Set) usually £94.99!).

Warnings

The authors strongly warn that you should “be extra critical and always consult a good mushroom book”, warning:

Warning: Never eat a mushroom because the system indicates you have found an edible fungus. Always seek advice from experts if you are not experienced yourself.

and

Note that the system should be used with great care, and not as tool to identify edible fungi without involving knowledgeable humans with experience in fungal recognition. So please explore with curiosity and sanity.

If users do follow the advice then great, I am concerned that they all won’t.

Tests

I installed and had a quick play with the Android version of the app (Danmarks svampeatlas). It has plenty of disclaimers about not eating them based on the identification. As expected, you take a photo and it gives you suggestions. You can then drill-down into each to see photos and description. At the time of writing it is still partly in Danish including the species names.

Far better, for now at least, is the web version of the same – https://svampe.databasen.org/imagevision

Upload a photo and it gives you suggestions. The species names are English, though not always the same common name as we widely use in the UK. You can drill down for their photo and details but that is in Danish, but along side several fungi books it’s very useable.

automated picture recognition on the Danish Fungal Atlas
Screenshot from the web version of the automated picture recognition on the Danish Fungal Atlas website. My uploaded Jelly Ear photo suggests a range of options, the first one being correct.

 

automated picture recognition on the Danish Fungal Atlas
Drilling down on the first suggestion gives their image and description, but in Danish.

Results

I did some tests with some of my many mushroom photos, edible, poisonous and non-edible species. Most identifications were correct (first species suggested) but of course will depend upon your photos / specimen:

Correct Species

  • Fly Agaric
  • Death Cap
  • Candle snuff
  • Parrot Waxcap
  • Scarlet Hood
  • Bay Bolete
  • Parasol (photo from underneath!)
  • Dryads Saddle
  • Field mushroom (picked specimen on its side)
  • Scarlet Elf Cups
  • Cauliflower Fungus
  • Chanterelles
  • Stinkhorn (even though the top was covered in flies!)
  • Bearded Tooth
  • Collared Earthstar
  • Scarlet Caterpillarclub
  • White Spindles

Correct to Genus (Group)

  • Royal Bolete
  • Holly Parachute Mushroom
  • Oak Bolete

Wrong Species and Genus

  • Sandy stilt puffball
  • Snowy Waxcap

I did try and push it a bit further (it’s a bit addictive!) and interestingly:

  • An icing Fly Agaric of my 50th birthday cake was correctly identified!
  • It identified Hedgehog mushrooms in a basket of them and Autumn Chanterelle (they were in separate groups).
  • Unsurprisingly, it failed on a very mixed basket.

Leave a Reply

Your email address will not be published. Required fields are marked *