Andrea Zampetti

PhD candidate in Animal Biology at @SapienzaRoma | Interests in #ConservationBiology, #EcologicalModelling, #CameraTraps, #AIforConservation

[Andrea Zampetti @aw_zampetti]

📢🧵 New paper out on @MethodsEcolEvol !

Can #MachineLearning be used to get accurate, unbiased density estimates in #cameratrap studies? Our findings suggest it can! 📷🦊 Check out how automated classification could improve wildlife monitoring!

https://t.co/PS0CiOWlUl

[Museo Nacional de Ciencias Naturales @mncn_csic]

🧵(HILO) Un equipo de investigación del MNCN-@CSIC ha desarrollado TropiCam-AI, el primer algoritmo de IA que identifica automáticamente, a partir de imágenes de cámaras trampa,las especies animales que habitan en el dosel de bosques húmedos neotropicales. https://t.co/IKofzVPyte

[Carlos Bautista @CBautistaLeon]

📢Wonderful postdoc opportunity!

Postdoc position in ecological modelling of camera trap data for biodiversity monitoring in the #Biodiversa WildIntel project https://t.co/lKlgggB0cJ

deadline: 1.09.2024

The position is affiliated with the @INCPoland.

https://t.co/rl8NvS82qs

[Methods in Ecology and Evolution @MethodsEcolEvol]

📖Published📖

Our new research article investigates whether foundation models are already performant enough to be used in ecology for animals' behavioural classification, from camera trap images, without any fine-tuning 📸 Find out more here 👇

https://t.co/5ml0e3KpOZ

[GMA Lab @global_mammal]

Excited to share that Martina Fernando, PhD student at our Global Mammal Assessment Lab, just presented her work "Developing a global probability map of illegal hunting on terrestrial mammals"with Michela Pacifici and Marco Davoli 🐘🌍#ConservationScience #IllegalHunting #ECR2025

[GMA Lab @global_mammal]

Finally published in Ecological Solutions and Evidence, paper titled "Ecologically sound strategies for renewable energy transition: Balancing conservation and energy development in coal plant conversions," by Michela Pacifici @global_mammal🌿
#GreenEnergy
https://t.co/XTWDNbVsVk

[nature @Nature]

Artificial intelligence is already used extensively to infer outcomes from tables of data, but this typically involves creating a model for each task. A one-size-fits-all model just made the process substantially easier https://t.co/pAtziaxq66

[cameratraps @cameratraps1]

Realy nice review on how do #ecologists estimate #occupancy in practice

It seems that single-species is the most common model family, #mammals the most common group, and unmarked #Rpackage the most common software

#cameratrapping

[Mark Gately @GatelyMark]

Wildlife photo of the week is an Anatolian leopard is caught by a camera trap set up in Ankara, Turkey. It’s thought there are fewer than 1,100 adult leopards left in the region.
https://t.co/ZAUw6AAJ2P

[cameratraps @cameratraps1]

First big step in automatic procedures using #AI to estimate population density using #cameratraps

@WildInsights and #MegaDetector as AI, and @DistanceSamp and REM for density. Simulations and fox, badger and roe deer as case study

By @aw_zampetti et al on @MethodsEcolEvol

[SCB Italy @scb_italy]

Siamo lieti di comunicarvi che stanno per arrivare i podcast "Volevo solo salvare i panda 🐼", una guida galattica per conservazionisti. L'intro al seguente link
https://t.co/zDRCmyZiCC

Stay tuned 😍

[Andrea Zampetti @aw_zampetti]

@jeremyjcusack @MethodsEcolEvol Thanks! Yes, to better isolate the effect of AI classification we manually extracted position and movement parameters, knowing that AI tools such as the one developed by @timmhaucke are already being tested in pratical applications: https://t.co/wk98UJ0y5E

[Andrea Zampetti @aw_zampetti]

📢🧵 New paper out on @MethodsEcolEvol !

Can #MachineLearning be used to get accurate, unbiased density estimates in #cameratrap studies? Our findings suggest it can! 📷🦊 Check out how automated classification could improve wildlife monitoring!

https://t.co/PS0CiOWlUl

[Andrea Zampetti @aw_zampetti]

@engmlubbad @MethodsEcolEvol Pretty excited to see in particular the development of LLMs work by the big research teams such as @MIT_CSAIL and @Microsoft #AIforGood initiative!

[Andrea Zampetti @aw_zampetti]

@engmlubbad @MethodsEcolEvol Hopefully in the immediate future we could see how more ecologist start to envision the potential of more automated approaches, and how this can benefit faster and more efficient research practices. ⤵️

[GMA Lab @global_mammal]

Pleased to announce that the latest paper is published in @consbiog by our postdoctoral researcher @MarcoDavoli3 with co-authors Michela Pacifici and Andrea Cristiano
@global_mammal @carlo_rondinini
#wildmammals #biodiversity #speciesrichness
https://t.co/XtiRu9tchM

[Diego J. Lizcano @dlizcano]

Towards an automated protocol for wildlife density estimation using camera‐traps - Zampetti - Methods in Ecology and Evolution - Wiley Online Library https://t.co/dJ4VrUyh1E

[Andrea Zampetti @aw_zampetti]

@sarameghanbeery @biorxivpreprint Many thanks, the paper is now also out on @MethodsEcolEvol!
https://t.co/NKThXFxmv0

[Andrea Zampetti @aw_zampetti]

📢🧵First paper out as a preprint on @biorxivpreprint!

📷🦊 We tested if machine learning for automated species classification can be used safely in camera-trap models for density estimation, and found out that this approach can yield unbiased estimates.
https://t.co/dirwaIgZuH

[Andrea Zampetti @aw_zampetti]

🫂 Thanks to the co-authors @Davide_Mirante, @cameratraps1 (Pablo Palencia) and @L_Santini_Eco for their critical contribution, and a special aknowledgement to the people at @TenutaStEgidio for their precious field support 🐗🦉! @SapienzaRoma @mncn_csic

[Andrea Zampetti @aw_zampetti]

‼️🔴 Our results suggest that popular algorithms for animal detection such as MegaDetector can be safely integrated with CT-DS and REM, and that final density estimates are reliable. Moreover, CT-DS showed to be robust even when taxonomic classifier accuracy was as low as 50%.