[john]

Thanks for the exciting update!  I’m going to begin diving into this new feature this week.  Are you guys based in SF?  I’d love to meet up with you all at some point.

[john]

Bump

[john]

Thanks for the quick response!

Question:

– Is there a timeline when this feature might be available?  This thread is nearly 3 months old now with the same response.  🙁

– Will scans/models created today be compatible with future versions of the SDK when this feature is rolled out?

 

Definitely am glad to hear there is a way to share scans.

 

Thanks!

• This reply was modified 8 years, 7 months ago by john.

[john]

Finally, I wanted to a build a global submission program that worked like this:

 

1) Aquire a sample (we’d start with powder only – no capsules or tablets).

2) Scan it with the SCIO given a well defined and reproducible set of  instructions of how to do it.

3) Anonymously mail the sample to the laboratory with an ID and publish the ID to a collaborative website.  Labrotory will publish the results with the ID from its gas chromatograph along with it’s opinion of what it is.

4) After discussion on the sample, if it’s pure, the sample is included in the training model.

 

Consumer physics: is there a way step 4 works with your platform?  We could solicit a massive amount of help from the community, both in terms of samples and donations to perform the laboratory scans.  My understanding is this would not work because there is no way to gain access to the raw scan data and share it with others.  Is this accurate?

• This reply was modified 8 years, 7 months ago by john.
• This reply was modified 8 years, 7 months ago by john.
• This reply was modified 8 years, 7 months ago by john.
• This reply was modified 8 years, 7 months ago by john.
• This reply was modified 8 years, 7 months ago by john.

[john]

Please see my notes in this thread -> https://dev.consumerphysics.com/forums/topic/mdma/

 

This endeavor fails feasibility until the SDK is capable of returning a confidence number of how well a scan matches a built model.  You can have the greatest model in the world built around MDMA, but if a user scans a banana the SDK is basically unable to tell the application that the provided sample does not match any of the loaded models.  It’s pointless to proceed until this is in place, unfortunately.

• This reply was modified 8 years, 7 months ago by john.
• This reply was modified 8 years, 7 months ago by john.

[john]

Hey Nick,

 

Yes, I have multiple developer kits to ensure that there does not exist any wild discrepancies between SCIO devices.

 

I am also disappointed there is no collaborative features to share scans with others in the community.  This endeavor will require participation in different locations around the country and world to be effective.  I’m not sure how we could collaborate on sample processing if consumer physics is hoarding all the scans and preventing collectors from sharing with one another.

• This reply was modified 8 years, 7 months ago by john.

[john]

Hey team,

 

I have access to a lab who can confirm the control groups when collecting these samples.  They can provide results from their gas spectrometer of any sample provided.  I have tested my application on over the counter Aspirin with some success.

 

However, all efforts surrounding harm reduction MUST NOT CONTINUE until Consumer Physics is capable of providing a confidence interval.  Trying to develop an application like this without such a metric available is reckless and nobody should attempt it until this is in place.

 

Consumer Physics:  how long until such a feature is ready?  What are some of the challenges you are encountering trying to develop this?  Unless I’m missing something here, I’d imagine pretty much most other applications cannot continue until the SDK is capable of returning a “no such sample found” response.

• This reply was modified 8 years, 7 months ago by john.

[john]

Thanks for the quick response.  I’ve been testing a classification model with some over the counter medication as a test and I’m optimistic with the results given a very poor model.

A detection threshold of even 1% would definitely prove to be useful for this use case.  The multitude of combinations that exist out there will definitely be problematic (but hey, that’s why we have a community to help!), but a more immediate and feasible use case would be to test if given sample is pure.  Any deviation from the well-tested model would throw an alert.

Have you guys made any more progress with returning a confidence threshold from a scan so a yes/no decisions can be made?  In the meantime, this application could have an “advanced view” that could show the user what the “ideal” spectra for pure MDMA powder is (we could ignore capsules and pills for now) and let the them decide if they want to further test the sample (with a chemical test, for example).

1) Is a confidence % against a model from a scan even feasible?

2) Also, will there come a point where we can share our models with each other in the community?  I’d imagine any such repositories would have rules in place needed to adhere to a strict standardized method of collecting data.

Thanks!

 

• This reply was modified 8 years, 9 months ago by john.
• This reply was modified 8 years, 9 months ago by john.
• This reply was modified 8 years, 9 months ago by john.

[Author]

Posts