Hi Dominique,

Thanks for your responses and congratulations on getting trained! 

I can see that speaking directly with a climbing professional could be the most beneficial because what climbing methods and equipment you may need will depend very much on individual project goals and budgets. Did you end up speaking with your trainers about your field research goals and what climbing methods may be best for you? 

Hi Mark, thanks for responding. I think you've identified one of the most difficult parts of research climbing: maintaining your climbing skills and knowledge between field sessions. 

My husband is an experienced arborist and practices his skills almost daily. I am not an arborist, so I schedule climbing time to keep my abilities fresh and my husband is there to assist. But I know it's difficult for individual researchers to practice on their own and they should only be climbing alone if experienced and not in a remote area.

However, it's possible to train researchers to safely climb in the field for multiple field sessions. My husband and I trained a group of climbers in Cameroon in January, 2024. The goal was to train four climbers who would go into the remote rainforest for several weeks and set up camera traps. They would deploy and retrieve arboreal cameras at different survey locations over two years. We needed to train the teams to operate safely and independently (without an instructor present) in very remote areas. 

To train them sufficiently, my husband and I spent 1 month in Cameroon with the field team. We did a few days of basic training at a location near the city and then went with them on their initial camera deployment where we continued field training for 2.5 - 3 weeks. Before going to Cameroon, we had several discussions to determine what climbing method and equipment would best meet their data collection goals and were appropriate for their field site and conditions. We taught them basic rescue scenarios. We also set a climbing practice schedule for the team to maintain their climbing and rescue skills. We strongly emphasized to their management that the field team needed access to the climbing gear and needed to follow the practice schedule. Since the training, the team successfully finished two other camera trap surveys and is planning their third.

This was a lot of up-front training and cost. However, these climbers are now operating on their own and can continue operating for the foreseeable future. I think a big reason is receiving extensive training, tailored to their needs. General tree-climbing courses are great for learning the basics, but they'll never be a substitute for in-field, tailored training.

Don't they all come from China ?

The one in the picture I can find on aliexpress for 2.53 euros.

I'm not sure how you would get one that doesn't come from China.

In any case, I know what to search for now, that's very helpful. Thank you.

What is the 1S thing you mention above ?

If you have any links as to what you consider a high quality one that would be great!

Yes, I have these. I'm pretty sure that my company is the first company in the world to sell products running on Raspberry Pi 5 in secure boot mode as well :-)

I responded to your wish list. From your description a modified camera like a HIKvision turret might be able to do it for you.

Thanks! The Teensys are nice for processing power if choosing an external Lora board I’d say that’s a good choice. I started with teensies, there was a well supported code base and certainly well priced.

My preference is to one with onboard Lora. I’ve used one one onboard murata Lora before. It was good both for low power operation and it’s Lora operation.

For a transmitter the grasshopper if it’s still being made is quite good for small distances are being made because it has an onboard ceramic antennae. Which is good for about three houses away, although one was also received > 20 km away.

Hello Patrick, thanks for asking! We are currently working on a bioacoustics module and will be releasing some time early next year. Maybe we can have some of your models in our initial bioacoustics model zoo, or if you don't have a model yet but have annotated datasets, we can probably train some models together? Let me know what you think! 

Thank you so much! We are also working on a bounding box based aerial animal detection model. Hopefully will release sometime early next year as well. It would be great to see how the model runs on your aerial datasets! We will keep you posted! 

That looks like an amazing project. Congratulations!

This is chatgpt's answer:

"Here’s a rough idea of the number of parameters in different types of models commonly used in Edge Impulse:

1. Neural Networks for Classification (e.g., MLP, CNN)
   • Small, shallow models: These can contain as few as 10,000 to 50,000 parameters when dealing with simple tasks like basic image recognition or sensor data classification.
   • Medium complexity models: For more advanced tasks, such as keyword spotting or more detailed image recognition, models may contain 50,000 to 200,000 parameters.
   • Larger models (Edge Optimized CNNs): For tasks like more complex object detection, models can go up to 200,000 to 500,000 parameters while still being optimized for edge devices.
2. Recurrent Neural Networks (RNN, LSTM)
   • These are often used for time-series data like accelerometer or audio signals. A typical RNN model for edge deployment might have 10,000 to 100,000 parameters, depending on the length and complexity of the sequences being analyzed.
3. Transfer Learning Models (e.g., MobileNet, EfficientNet variants)
   • Transfer learning models like MobileNet, when adapted for edge deployment, are typically pruned and quantized to reduce their size. These models may contain anywhere from 1 million to 3 million parametersin their full form, but when trimmed for edge deployment, they could be reduced to 100,000 to 1 million parameters.
4. Classical Machine Learning Models (e.g., Decision Trees, KNN)
   • These models usually don’t have parameters in the same way as deep learning models, but they are lightweight and typically optimized for embedded use. They might only use a few kilobytes of memory or less, depending on the dataset size and number of features.

Overall, the model sizes in Edge Impulse are kept small to run efficiently on microcontrollers with limited memory (usually in the range of 256 KB to 1 MB of RAM and a few megabytes of flash). The number of parameters is carefully balanced to ensure good accuracy while maintaining small memory footprints for edge deployment."

Disclaimer: I can't guarantee that all of the above is not a hallucination.

So, not even 1 million parameters. yolov3 tiny model back in the day was 6 million parameters. It's my understanding that the idea behind edge impulse was that it's retrained up at the edge itself. Training requires a lot of memory than inference so by it's nature the models would be further constrained by this.

It's important to be aware of what your model is (How advanced it is), what size it is and how many parameters it contains when you consider your use case to manage your expectations. Later I will post you a few of the terrrible false positives that even the very best and largest models have made to give you an idea of just what AI is actually learning and by extension not learning. So if the best of the best can make terrible mistakes, what will small early models do?

In general though, a large advanced model can do much better at inference with a relatively small number of annotated training images than a really small model with potentially thousands and many more images.

For example, I decided way back in 2019 that yolov3 tiny model was absolutely useless for security because of the boy who cried wolf effect. 
 

The biggest use case I see for very small models would be as a hopefully better replacement to using a PIR sensor. I say hopefully, because both PIR sensors and very small models can have de-generate environmental conditions that cause large number of false positives. At a certain threshold, any perceived gain in power usage may disappear.

I welcome edge impulse to correct anything above that I may have said that was untrue as I personally have never used edge impulse, but I was involved in a project whereby yolov6 was pitted against and compared against a system using edge impulse.

Yes baboons. But this area has a few different primate species. In the weekend three species were present.

Yes, the thermal ones will be spectacular. I wondering what other species will jump into the picture when we start with that. The 640x512 thermal camera I'm testing is really amazing. Now I'm onto testing a 1280x1024 resolution thermal camera. That's going to be super interesting! As soon as I have some nice images I'll post them.

Sure, I'd like to , but don't have the badge yet. ;) 

Hi Carly,

I've got you point, that's the reason why we want to make a new edition of our Planet Watch initiative to support the scientific community.

Best regards,

Adrien