Render image volumes for bias investigation

[wrgr]

Render N volumes with low density and N with high density for human assessment to check for bias.
Will human label (without detections, but with masks) as high or low density in future weeks across ~200 volumes)

[jovo]

:ok_hand:

On Wed, Jul 27, 2016 at 9:34 AM, William Gray Roncal < notifications@github.com> wrote:

Render N volumes with low density and N with high density for human assessment to check for bias.

Will human label (without detections, but with masks) as high or low density in future weeks across ~200 volumes)

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/neurodata/ndpaper/issues/79, or mute the thread https://github.com/notifications/unsubscribe-auth/AACjcgOV6vhadPrye2ogpn3xJ35viByvks5qZ15mgaJpZM4JWMZD .

the glass is all full: half water, half air. neurodata.io, jovo calendar https://calendar.google.com/calendar/embed?src=joshuav%40gmail.com&ctz=America/New_York

[wrgr]

Almost there on this - turned out to be hard.

I have gifs rendering for arbitrary clips and numpy arrays - but having trouble with rendering with overlays. Think I've found a good way around this and a backup plan just in case.

Estimated completion: Friday 8/5.

[jovo]

ok

On Thu, Aug 4, 2016 at 6:10 AM, William Gray Roncal < notifications@github.com> wrote:

Almost there on this - turned out to be hard.

I have gifs rendering for arbitrary clips and numpy arrays - but having trouble with rendering with overlays. Think I've found a good way around this and a backup plan just in case.

Estimated completion: Friday 8/5.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/neurodata/ndpaper/issues/79#issuecomment-237511442, or mute the thread https://github.com/notifications/unsubscribe-auth/AACjchjF-WmwxRfTzl6eeRadmsMqigErks5qcbqCgaJpZM4JWMZD .

the glass is all full: half water, half air. neurodata.io, jovo calendar https://calendar.google.com/calendar/embed?src=joshuav%40gmail.com&ctz=America/New_York

[wrgr]

can now make these volumes, so closing this issue. we need to talk through exactly how many and how to blind. i will proceed with my best idea if we don't get a chance to chat.

[jovo]

i thought we said 100 of each? though open to other suggestions.

On Sun, Aug 7, 2016 at 9:40 PM, William Gray Roncal < notifications@github.com> wrote:

Closed #79 https://github.com/neurodata/ndpaper/issues/79.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/neurodata/ndpaper/issues/79#event-747838065, or mute the thread https://github.com/notifications/unsubscribe-auth/AACjckwgLeu3TWeO7pErADz_lJiR66N1ks5qdokMgaJpZM4JWMZD .

the glass is all full: half water, half air. neurodata.io, jovo calendar https://calendar.google.com/calendar/embed?src=joshuav%40gmail.com&ctz=America/New_York

[wrgr]

just want to be explicit:

• 100 volumes of each mean +/- 1 stddev
• do in a way that presentation is blind to will, which might mean someone else generates labels or will figures out how to do where he can't see values (ever).
• will creates a 200x1 vector of high/low values (binary)
• joshua or someone evaluates and sees how good will is at this
• we decide how to proceed

yah?

[jovo]

good start. WWW: who, what, by when. we got who & what. by when though?

On Sun, Aug 7, 2016 at 11:09 PM, William Gray Roncal < notifications@github.com> wrote:

just want to be explicit:

• 100 volumes of each mean +/- 1 stddev
• do in a way that presentation is blind to will, which might mean someone else generates labels or will figures out how to do where he can't see values (ever).
• will creates a 200x1 vector of high/low values (binary)
• joshua or someone evaluates and sees how good will is at this
• we decide how to proceed

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/neurodata/ndpaper/issues/79#issuecomment-238132627, or mute the thread https://github.com/notifications/unsubscribe-auth/AACjcqTpEdAqyzxHpOUDbf8Ql32BqnYVks5qdp3PgaJpZM4JWMZD .

the glass is all full: half water, half air. neurodata.io, jovo calendar https://calendar.google.com/calendar/embed?src=joshuav%40gmail.com&ctz=America/New_York