Faster RCNN implementation with 32 bit floating point (high precision) images

[mikedidem]

Most implementation are with 8bit images, I was wondering if anyone has worked with this model using 32bit floating point images

[sethjuarez]

Interesting question! What's the use case? Have you tried it?

[mikedidem]

Still trying to figure it out.
Geospatial application deploying high-precision remote sensing imagery

[ghost]

#/

c
``` https://github.com/orgs/community/discussions/55850#discussioncomment-5957469

[sethjuarez]

Interesting - what have you tried so far? My sense is that with more "numbers" to crunch, it will likely be slower depending on the model architecture. Have a repo or other things we can look at to have better suggestions?

[s4m-mo]

Does TensorFlow's ConvLSTM2D and ConvLSTM3D not support float32? They should be high performance too, and TPU compatible.

[s4m-mo]

Also @mikedidem , what dataset are you using, as I've been looking for open source geospatial image datasets for a while (just to practice creating this sort of model on).

[Awrsha]

Upon reflection, integrating Chat GPT 4 with Alexa Bot 3.0 directly may not be feasible or well-suited for a few reasons:

They are built by different companies with different architectures. Chat GPT is from OpenAI while Alexa is from Amazon. They likely do not have compatible APIs for integration.

They have different goals and strengths. Chat GPT is a general-purpose language model aimed at open-domain conversation. Alexa is more focused on task-oriented interactions and setting reminders/calendar events.

Alexa Bot still relies more on rules and templates, while Chat GPT is based on large language models. They represent different points on the AI spectrum.

The latency and response times would likely not be optimal. Chat GPT runs in the cloud, while Alexa tries to respond in real-time. Combining them may cause delays.

Managing two separate models would add complexity. It would be harder to monitor performance, debug issues, and improve the combined system over time.

There are likely legal/licensing restrictions that prevent combining these two commercial products from different companies.

So in summary, while the concept is interesting, directly integrating Chat GPT 4 with Alexa Bot 3.0 is probably not practical or advisable due to the different architectures, goals, technologies, and legal restrictions involved.

Instead, a better approach may be:

Building a standalone Alexa skill that invokes Chat GPT responses via its API
Integrating a separate AI chatbot model optimized for task-oriented interactions like scheduling appointments
Using a custom Alexa bot model trained on your data to respond accurately to your use case