@janardhanp22 and I can summarize the talks. We were initially thinking of doing detailed writeups of each project along with the speaker as it is hard for people to get involved without sufficient details. Especially, the more accurate domain knowledge that we can share we will be enabling more people to contribute/ collaborate. Even better if we can include concrete AI tasks, benchmarks, and data sets. But I guess, we can do summary followed by series of blog posts, as I would imagine detailed posts would take time.
Regarding the seminar series, I think that would be great! I spoke to Prof. Yannet Interian during the class and she mentioned that May 5th is open and she blocked it tentatively for our talks. Seems like it would be 1:30 hour session with 1 hour talks and 30 minutes for Q&A? Are you thinking of one session or multiple sessions?
As it is just 1 hr slot for speaking, I was thinking we might have to split the talks into a couple sessions? That is, assuming everyone who spoke is willing to speak in the seminar series. Also, we do have 2-3 speakers lined up for the next slot. Hopefully, we will have more ideas coming in too.
These talks are very inspiring, I could spend only less time in the last 2 weeks and got little distracted due to work. Thanks to @sravya8 . I am very interested in identifying plant diseases using Deep Learning. Currently going through this paper https://arxiv.org/abs/1604.03169v2 to see if I can get something for a talk. Surprisingly this is the only paper I could find related to this area.
Thanks @renjithmadhavan ! @singlasahil14 uses satellite imagery to analyze crops, may be you both can talk And do add yourself into the speakers list. Just mention that you would update with the date later if you are not ready yet.
@renjithmadhavan, plant disease image recognition is an interesting space. I played with plant disease in part 1 with vgg16 and vgg16bn and got good results. I’ve also seen folks using resnet with much better outcome though you need to train a model from scratch.
I believe Jeremy showed another paper in lesson 1 slides, I don’t have the link handy though. As far as I know, there are two groups looking into this, plantvillage+crowdai and plantix+peat, the former focusing on crowdsourcing ai, the later focusing on crowdsourcing data. I think peat works with farming community in India and got some really good work done there.
Thats great. Right now I am trying to build a android apps ( really a hello world ). I plan to start with a pre-trained model. But my focus is on getting a useful product(like an android prototype quickly).
This looks great! Thanks so much for putting it together.
I just have one high level comment - I think you’re significantly underselling the extremely impressive results and methods shown in the talks. To think that just a few months ago these are folks who had no prior deep learning background (and in many cases, little or no machine learning background!) and (on the whole) no computer science qualifications, it’s just amazing and fantastic to see what you’ve all achieved! (And all just part-time students at that…)
This is important to me because it’s the strongest possible validation of the claim that deep learning need not be the exclusive domain of the PhD graduates of a few famous schools, but can be used by anyone prepared to invest a few months of evenings and weekends. I’d hope it would be very encouraging and validating to folks that are on the outside looking in, and wondering if they have any chance of success themselves…
Added some more points at the end as you mentioned. Let me know if we were able to do justice in terms of wording
We were also thinking of naming the article “What is the biggest positive impact we could have with AI?” as that has been our uber-ambitious goal for proposing this talks. How does that sound or is it too click baity?
Thanks @jeremy for your valuable comments. It echoed the point where most of the folks are just rejected without the chance to prove themselves just because of the fact that they don’t have a strong Math or Phd background from very famous schools.
@sara.hooker13 FYI, the summary has been published, feel free to share it
Also, with respect to calculating the direction/distance in the Rainforest project, I was thinking having a way to rotate the Guardian(cell phone) around the trunk of the tree when it detects chainsaw might give us information to calculate the direction and probably distance too? I know it sounds too fancy but does not seem impossible
And do add yourself into the speakers list. Just mention that you would update with the date later if you are not ready yet.