Improving object detection colab #240
Conversation
Giom-V
commented
Jul 11, 2024
•
Giom-V
commented
- fixing the 4th example by removing the count and the parsing that was depending on Gemini's output
- Adding a comment about lowering the safety features (to be discussed if it works without lowering the safety features)
- Simplifying the Open vocabulary object detection part
* fixing the 4th exemple by removing the count and the parsing that was depending on Gemini's output * Adding a comment about lowerering the safety features
…ards other examples.
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
github-actions
bot
added
status:awaiting review
…s) and grouping both parts with the castle drawing.
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-08-28T06:28:52Z Can you change the references to "Gemini" to say "Gemini API"? Gemini is the consumer-facing chatbot, it's confusing, so we need to be clear (it's something marketing legal request we do).
(as a side note, I wonder if we could use structured JSON output schemas to make the output parsing simpler and stable. just a thought, not something for this change) Giom-V commented on 2024-09-11T12:46:11Z Done for the references to Gemini. I will think of structured output, it's a good idea. Giom-V commented on 2024-09-11T13:53:04Z For the structured output, since this PR is already complicated, I propose to do it in a second one if that's ok with you. markmcd commented on 2024-09-12T08:09:47Z do it in a second one if that's ok with you.
yeah for sure - sorry, I was just thinking out loud! |
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-08-28T06:28:52Z Typo: |
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-08-28T06:28:53Z Line #2. # Adapted from https://colab.research.google.com/drive/1bjGbrtjE_Ugc3YpIvQMtkqsiMPqIp89W#scrollTo=02GJS3Zv4JAu written by Andreas Steiner and Gabriel Antoine le Roux
If this work is derived from that link, you'll need to include their copyright line in the header at the top of this file along with ours. Giom-V commented on 2024-09-11T12:51:41Z Do you have an example for that? I haven't changed this part of the code but I agree we should do it right... markmcd commented on 2024-09-12T08:21:00Z It may be sufficient to just include the copyright line from that Drive URL at the top of our file but I'll send an email to people who can answer with authority. I'll CC you. |
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-08-28T06:28:54Z Nit: Giom-V commented on 2024-09-11T12:52:10Z Fixed |
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-08-28T06:28:55Z Line #1. url = "https://github.com/geminidemospatial/SpatialDemo/blob/main/gem_dogs.png?raw=true"
What are these images? There's no license in the repo, so the owner retains copyright, and we are not permitted to distribute them (e.g. by embedding them in a notebook and uploading to github)
If they are a Googler you could reach out and ask them to document the repo properly so we can use it? Giom-V commented on 2024-09-11T12:54:14Z I have absolutely no idea who it is, so I guess I'll have to change the pictures. Giom-V commented on 2024-09-11T13:53:25Z Replaced with a new picture from wikipedia. |
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-08-28T06:28:55Z Nits:
Giom-V commented on 2024-09-11T12:55:22Z Fixed |
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-08-28T06:28:56Z If we're distributing these images (e.g. by including them in a notebook we upload to GitHub), we need to comply with their license (which appears to be MIT?). Giom-V commented on 2024-09-11T12:56:46Z Not sure what it entails to be honest. I'll talk to you directly about that. |
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-08-28T06:28:57Z Nit: Giom-V commented on 2024-09-11T12:56:53Z Fixed |
|
Left some review feedback - mostly little nits but there are a couple of licensing questions we need to answer too. |
markmcd
added
status:awaiting response
|
Done for the references to Gemini. I will think of structured output, it's a good idea. View entire conversation on ReviewNB |
|
Do you have an example for that? I haven't changed this part of the code but I agree we should do it right... View entire conversation on ReviewNB |
|
Fixed View entire conversation on ReviewNB |
|
I have absolutely no idea who it is, so I guess I'll have to change the pictures. View entire conversation on ReviewNB |
|
Fixed View entire conversation on ReviewNB |
|
Not sure what it entails to be honest. I'll talk to you directly about that. View entire conversation on ReviewNB |
|
Fixed View entire conversation on ReviewNB |
* removing interactive section to simplify
|
For the structured output, since this PR is already complicated, I propose to do it in a second one if that's ok with you. View entire conversation on ReviewNB |
|
Replaced with a new picture from wikipedia. View entire conversation on ReviewNB |
do it in a second one if that's ok with you.
yeah for sure - sorry, I was just thinking out loud! View entire conversation on ReviewNB |
|
It may be sufficient to just include the copyright line from that Drive URL at the top of our file but I'll send an email to people who can answer with authority. I'll CC you. View entire conversation on ReviewNB |
|
View / edit / reply to this conversation on ReviewNB markmcd commented on 2024-09-12T08:29:30Z This cell has a weird error output, and the next one has an error too. Make sure they get cleaned up before merging. Giom-V commented on 2024-09-13T09:42:34Z I updated the outputs. It should be better now. |
|
I updated the outputs. It should be better now. View entire conversation on ReviewNB |
Giom-V
requested review from
markmcd and
MarkDaoust
and removed request for
MarkDaoust
