Chimp&See February 2026 Newsletter

Dear Chimp&See-ers,

Time passes so quickly when you’re having fun! We hope you all have had a wonderful start to 2026!

Our last newsletter was in October, and since then a lot has happened:

• We finished the Eternal Victory site SpeciesID and TrottersID workflows (MonkeySee still needing classifications)

• We opened and closed the Cascasing Arches site SpeciesID (TrottersID for CA up now!)

• We opened a new site in January called Crystal Gully. This site is located in Gabon in a more savannah-y area, so lots of new things to see and discover - like the first lion seen in Gabon since the 90s! — so come join us at Crystal Gully SpeciesID next time you are online!

You may be wondering how we have been getting through sites so much faster than before, and that is thanks to the assistance of our AI platform Zamba!

Zamba is a machine learning algorithm the PanAf developed with DrivenData to automate species identification from camera trap videos, and it's all based on Chimp&See annotations!

Because even Zamba makes errors, we can’t use it by itself quite yet. But since it gets things right most of the time, we have been combining its answers with our Chimp&See community science annotations for the past few sites to reduce the number of viewers needed to retire a clip. You can read more about how we are doing that in Colleen’s post on Talk.

All this efficiency with SpeciesID has caused a bit of a backlog on MonkeySee, TrottersID, and chimp matching, and that is A-OK. We will take our time with those, BUT if you want to help, please join those side quests!

We also have noticed fewer tags since fewer people are viewing each video, so please feel free to tag as you classify when you are very sure of a species to try to bring the number up to past levels.

Wishing you all the very best for 2026, and see you at Chimp&See.org!

Cheers,

Mimi & the Chimp&See team🐵

 

Facebook AI/PanAf collaboration update - Densepose: AI for orienting humans (and chimps) in space

We're happy to announce the results from our first collaboration with facebook AI where they used the PanAf chimpanzee videos to test whether their Densepose algorithm could retrain itself to identify the planes of a new species. Originally Densepose was trained on human videos and can identify an individual's horizontal and vertical planes - so it can identify what is the front and the back, the left side and right side, and  the top and bottom of various body parts, which all move together to document how a person (or chimp) is moving.

In the example below you can see:
Top left panel - a still from the original video with 2 chimps
Top right panel - each of the body parts that can be detected by Densepose, head, hand, forearm, upper arm, torso, etc each coded in a different colour.
Lower left panel - the horizontal plane is coded for each body part. For example, the left side of the head is yellow, the right side is blue.
Lower right pane - the vertical plane is coded: the top of the head is yellow and the bottom is blue.

Put all together, the animal can be oriented in space!

For more information, check out their blog post:  https://ai.facebook.com/blog/dense-pose-for-animal-classes-using-transfer-learning