added results data and plotting nb for paper figures

[ulisrael]

I've added all the data we generated necessary for generating plots that are found in the paper

• eval_results
• livecell
• metadata_LIVECELL

Added a notebook that will generate each result in the paper.

[rossbar]

Thanks @ulisrael - one thing immediately jumps out:

There appear to be some hard-coded paths in the notebook so it's not runnable. Could you go through the following exercise to ensure that the script is runnable and the results are reproducible:
1. Freshly clone the cellsam repo
2. Checkout this branch (git checkout --track upstream/figures)
3. Create a new virtual environment - since this is on master you'll need to use 3.10, so something like python3.10 -m venv cellsam-dev then source cellsam-dev/bin/activate.
4. Install all the dependencies pip install -r requirements.txt and cellsam: pip install .
5. Run the notebook in the way that you usually do to generate the figures

Once the figure gen is runnable, then we can move onto the next steps. Some impressions I have are:

1. I didn't realize the results would be spread across so many files. Given that, I think it'd be best to store them outside the repo (e.g. in a bucket) to prevent tracking them.
2. Before merging, we should also be sure that the evaluation component produces a results "package" that is runnable by this same script (e.g. the directory structure and filenames are consistent, etc.)

[rossbar]

I took the liberty of pushing up most of the changes I mentioned above - make sure to pull them before pushing anything new here!

This still shouldn't be merged until we have a handle on the evaluation pipeline that generates these data, so IMO any spare bandwidth should go to the eval pipeline!

[ulisrael]

lgtm - I slightly modified the paths so that each section can be run independently without having to run the first file.

[ulisrael]

I've added all the data we generated necessary for generating plots that are found in the paper

• eval_results
• livecell
• metadata_LIVECELL

Added a notebook that will generate each result in the paper.

[rossbar]

Thanks @ulisrael - one thing immediately jumps out:

There appear to be some hard-coded paths in the notebook so it's not runnable. Could you go through the following exercise to ensure that the script is runnable and the results are reproducible:
1. Freshly clone the cellsam repo
2. Checkout this branch (git checkout --track upstream/figures)
3. Create a new virtual environment - since this is on master you'll need to use 3.10, so something like python3.10 -m venv cellsam-dev then source cellsam-dev/bin/activate.
4. Install all the dependencies pip install -r requirements.txt and cellsam: pip install .
5. Run the notebook in the way that you usually do to generate the figures

Once the figure gen is runnable, then we can move onto the next steps. Some impressions I have are:

1. I didn't realize the results would be spread across so many files. Given that, I think it'd be best to store them outside the repo (e.g. in a bucket) to prevent tracking them.
2. Before merging, we should also be sure that the evaluation component produces a results "package" that is runnable by this same script (e.g. the directory structure and filenames are consistent, etc.)

[rossbar]

I took the liberty of pushing up most of the changes I mentioned above - make sure to pull them before pushing anything new here!

This still shouldn't be merged until we have a handle on the evaluation pipeline that generates these data, so IMO any spare bandwidth should go to the eval pipeline!

[ulisrael]

lgtm - I slightly modified the paths so that each section can be run independently without having to run the first file.

[ulisrael]

I've added all the data we generated necessary for generating plots that are found in the paper

• eval_results
• livecell
• metadata_LIVECELL

Added a notebook that will generate each result in the paper.

[rossbar]

Thanks @ulisrael - one thing immediately jumps out:

There appear to be some hard-coded paths in the notebook so it's not runnable. Could you go through the following exercise to ensure that the script is runnable and the results are reproducible:
1. Freshly clone the cellsam repo
2. Checkout this branch (git checkout --track upstream/figures)
3. Create a new virtual environment - since this is on master you'll need to use 3.10, so something like python3.10 -m venv cellsam-dev then source cellsam-dev/bin/activate.
4. Install all the dependencies pip install -r requirements.txt and cellsam: pip install .
5. Run the notebook in the way that you usually do to generate the figures

Once the figure gen is runnable, then we can move onto the next steps. Some impressions I have are:

1. I didn't realize the results would be spread across so many files. Given that, I think it'd be best to store them outside the repo (e.g. in a bucket) to prevent tracking them.
2. Before merging, we should also be sure that the evaluation component produces a results "package" that is runnable by this same script (e.g. the directory structure and filenames are consistent, etc.)

[rossbar]

I took the liberty of pushing up most of the changes I mentioned above - make sure to pull them before pushing anything new here!

This still shouldn't be merged until we have a handle on the evaluation pipeline that generates these data, so IMO any spare bandwidth should go to the eval pipeline!

[ulisrael]

lgtm - I slightly modified the paths so that each section can be run independently without having to run the first file.

[ulisrael]

I've added all the data we generated necessary for generating plots that are found in the paper

• eval_results
• livecell
• metadata_LIVECELL

Added a notebook that will generate each result in the paper.

[rossbar]

Thanks @ulisrael - one thing immediately jumps out:

There appear to be some hard-coded paths in the notebook so it's not runnable. Could you go through the following exercise to ensure that the script is runnable and the results are reproducible:
1. Freshly clone the cellsam repo
2. Checkout this branch (git checkout --track upstream/figures)
3. Create a new virtual environment - since this is on master you'll need to use 3.10, so something like python3.10 -m venv cellsam-dev then source cellsam-dev/bin/activate.
4. Install all the dependencies pip install -r requirements.txt and cellsam: pip install .
5. Run the notebook in the way that you usually do to generate the figures

Once the figure gen is runnable, then we can move onto the next steps. Some impressions I have are:

1. I didn't realize the results would be spread across so many files. Given that, I think it'd be best to store them outside the repo (e.g. in a bucket) to prevent tracking them.
2. Before merging, we should also be sure that the evaluation component produces a results "package" that is runnable by this same script (e.g. the directory structure and filenames are consistent, etc.)

[rossbar]

I took the liberty of pushing up most of the changes I mentioned above - make sure to pull them before pushing anything new here!

This still shouldn't be merged until we have a handle on the evaluation pipeline that generates these data, so IMO any spare bandwidth should go to the eval pipeline!

[ulisrael]

lgtm - I slightly modified the paths so that each section can be run independently without having to run the first file.

[ulisrael]

I've added all the data we generated necessary for generating plots that are found in the paper

• eval_results
• livecell
• metadata_LIVECELL

Added a notebook that will generate each result in the paper.

[rossbar]

Thanks @ulisrael - one thing immediately jumps out:

There appear to be some hard-coded paths in the notebook so it's not runnable. Could you go through the following exercise to ensure that the script is runnable and the results are reproducible:
1. Freshly clone the cellsam repo
2. Checkout this branch (git checkout --track upstream/figures)
3. Create a new virtual environment - since this is on master you'll need to use 3.10, so something like python3.10 -m venv cellsam-dev then source cellsam-dev/bin/activate.
4. Install all the dependencies pip install -r requirements.txt and cellsam: pip install .
5. Run the notebook in the way that you usually do to generate the figures

Once the figure gen is runnable, then we can move onto the next steps. Some impressions I have are:

1. I didn't realize the results would be spread across so many files. Given that, I think it'd be best to store them outside the repo (e.g. in a bucket) to prevent tracking them.
2. Before merging, we should also be sure that the evaluation component produces a results "package" that is runnable by this same script (e.g. the directory structure and filenames are consistent, etc.)

[rossbar]

I took the liberty of pushing up most of the changes I mentioned above - make sure to pull them before pushing anything new here!

This still shouldn't be merged until we have a handle on the evaluation pipeline that generates these data, so IMO any spare bandwidth should go to the eval pipeline!

[ulisrael]

lgtm - I slightly modified the paths so that each section can be run independently without having to run the first file.

[ulisrael]

I've added all the data we generated necessary for generating plots that are found in the paper

• eval_results
• livecell
• metadata_LIVECELL

Added a notebook that will generate each result in the paper.

[rossbar]

Thanks @ulisrael - one thing immediately jumps out:

There appear to be some hard-coded paths in the notebook so it's not runnable. Could you go through the following exercise to ensure that the script is runnable and the results are reproducible:
1. Freshly clone the cellsam repo
2. Checkout this branch (git checkout --track upstream/figures)
3. Create a new virtual environment - since this is on master you'll need to use 3.10, so something like python3.10 -m venv cellsam-dev then source cellsam-dev/bin/activate.
4. Install all the dependencies pip install -r requirements.txt and cellsam: pip install .
5. Run the notebook in the way that you usually do to generate the figures

Once the figure gen is runnable, then we can move onto the next steps. Some impressions I have are:

1. I didn't realize the results would be spread across so many files. Given that, I think it'd be best to store them outside the repo (e.g. in a bucket) to prevent tracking them.
2. Before merging, we should also be sure that the evaluation component produces a results "package" that is runnable by this same script (e.g. the directory structure and filenames are consistent, etc.)

[rossbar]

I took the liberty of pushing up most of the changes I mentioned above - make sure to pull them before pushing anything new here!

This still shouldn't be merged until we have a handle on the evaluation pipeline that generates these data, so IMO any spare bandwidth should go to the eval pipeline!

[ulisrael]

lgtm - I slightly modified the paths so that each section can be run independently without having to run the first file.

[ulisrael]

I've added all the data we generated necessary for generating plots that are found in the paper

• eval_results
• livecell
• metadata_LIVECELL

Added a notebook that will generate each result in the paper.

[rossbar]

Thanks @ulisrael - one thing immediately jumps out:

There appear to be some hard-coded paths in the notebook so it's not runnable. Could you go through the following exercise to ensure that the script is runnable and the results are reproducible:
1. Freshly clone the cellsam repo
2. Checkout this branch (git checkout --track upstream/figures)
3. Create a new virtual environment - since this is on master you'll need to use 3.10, so something like python3.10 -m venv cellsam-dev then source cellsam-dev/bin/activate.
4. Install all the dependencies pip install -r requirements.txt and cellsam: pip install .
5. Run the notebook in the way that you usually do to generate the figures

Once the figure gen is runnable, then we can move onto the next steps. Some impressions I have are:

1. I didn't realize the results would be spread across so many files. Given that, I think it'd be best to store them outside the repo (e.g. in a bucket) to prevent tracking them.
2. Before merging, we should also be sure that the evaluation component produces a results "package" that is runnable by this same script (e.g. the directory structure and filenames are consistent, etc.)

[rossbar]

I took the liberty of pushing up most of the changes I mentioned above - make sure to pull them before pushing anything new here!

This still shouldn't be merged until we have a handle on the evaluation pipeline that generates these data, so IMO any spare bandwidth should go to the eval pipeline!

[ulisrael]

lgtm - I slightly modified the paths so that each section can be run independently without having to run the first file.

[ulisrael]

I've added all the data we generated necessary for generating plots that are found in the paper

• eval_results
• livecell
• metadata_LIVECELL

Added a notebook that will generate each result in the paper.

[rossbar]

Thanks @ulisrael - one thing immediately jumps out:

There appear to be some hard-coded paths in the notebook so it's not runnable. Could you go through the following exercise to ensure that the script is runnable and the results are reproducible:
1. Freshly clone the cellsam repo
2. Checkout this branch (git checkout --track upstream/figures)
3. Create a new virtual environment - since this is on master you'll need to use 3.10, so something like python3.10 -m venv cellsam-dev then source cellsam-dev/bin/activate.
4. Install all the dependencies pip install -r requirements.txt and cellsam: pip install .
5. Run the notebook in the way that you usually do to generate the figures

Once the figure gen is runnable, then we can move onto the next steps. Some impressions I have are:

1. I didn't realize the results would be spread across so many files. Given that, I think it'd be best to store them outside the repo (e.g. in a bucket) to prevent tracking them.
2. Before merging, we should also be sure that the evaluation component produces a results "package" that is runnable by this same script (e.g. the directory structure and filenames are consistent, etc.)

[rossbar]

I took the liberty of pushing up most of the changes I mentioned above - make sure to pull them before pushing anything new here!

This still shouldn't be merged until we have a handle on the evaluation pipeline that generates these data, so IMO any spare bandwidth should go to the eval pipeline!

[ulisrael]

lgtm - I slightly modified the paths so that each section can be run independently without having to run the first file.