Yesterday I came across this news and was deeply saddened by the fact that so many farmers are committing suicides and we're not able to help them out. So, I asked this question to myself (and my friends), that what can we, as Engineers, do to solve this problem. This repo is a log of fruitful discussions I had about this and will hopefully lead to something substantial, someday :).
Note 1: I understand that to solve this problem, at scale, Govt. needs to do lots of things, and we have to believe/accept that the Govt. is doing what it can do (otherwise we should be in politics), So, we can't crib about Govt./Country/Population etc (atleast on this repo). These are constraints to our optimization problem!
Note 2: Contributions of Ideas/References/Links to Agricultural-Datasets/<Anything else which you feel might be useful for engineers working in agricultural domain> etc are most welcome on this repo.
| Organization/Dataset | Description of dataset | Source |
|---|---|---|
| Image-Net Dataset | Images of various plants (trees, vegetables, flowers) | http://image-net.org/explore?wnid=n07707451 |
| ImageNet Large Scale Visual Recognition Challenge (ILSVRC) | Images that allow object localization and detection | http://image-net.org/challenges/LSVRC/2017/#det |
| University of Arkansas, Plants Dataset | Herbicide injury image database | https://plants.uaex.edu/herbicide/ http://www.uaex.edu/yard-garden/resource-library/diseases/ |
| EPFL, Plant Village Dataset | Images of various crops and their diseases | https://www.plantvillage.org/en/crops |
| Leafsnap Dataset | Leaves from 185 tree species from the Northeastern United States. | http://leafsnap.com/dataset/ |
| LifeCLEF Dataset | Identity, geographic distribution and uses of plants | http://www.imageclef.org/2014/lifeclef/plant |
| PASCAL Visual Object Classes Dataset | Images of various animals (birds, cats,horses, sheep etc. cows, dogs, | http://host.robots.ox.ac.uk/pascal/VOC/ |
| Africa Soil Information Service (AFSIS) dataset | Continent-wide digital soil maps for Sub-Saharan Africa | http://africasoils.net/services/data/ |
| UC Merced Land Use Dataset | A 21 class land use image dataset | http://vision.ucmerced.edu/datasets/landuse.html |
| MalayaKew Dataset | Scan-like images of leaves from 44 species classes. | http://web.fsktm.um.edu.my/~cschan/downloads_MKLeaf_dataset.html |
| Crop/Weed Field Image Dataset | Field images, vegetation segmentation masks and crop/weed plant type annotations. | https://github.com/cwfid/dataset https://pdfs.semanticscholar.org/58a0/9b1351ddb447e6abdede7233a4794d538155.pdf |
| University of Bonn Photogrammetry, IGG | Sugar beets dataset for plant classification as well as localization and mapping. | http://www.ipb.uni-bonn.de/data/ |
| Flavia leaf dataset | Leaf images of 32 plants. | http://flavia.sourceforge.net/ |
| Syngenta Crop Challenge 2017 | 2,267 of corn hybrids in 2,122 of locations between 2008 and 2016, together with weather and soil conditions | https://www.ideaconnection.com/syngenta-crop-challenge/challenge.php |
Reference:
Research papers:
- https://arxiv.org/ftp/arxiv/papers/1807/1807.11809.pdf
- Deep Learning in agriculture
- Research Articles
- A tool/platform which can raise funds in case of drought/when farmers are in dire need of money - because of which they eventually commit suicide - like paytm etc do in case of floods, earthquake or any other natural calamity. We need to build something like this especially for farmers, we need to have a mediator, backed by technology, which can directly connect needy farmers and those who can donate. We need an organization to maintain this platform's credibility and to run it!
- Generate passive sources of income, so that agriculture is not the only source of income for farmers. They should be able to participate remotely in that activity. One such activity - is dataset collection - but this isn't sustainable. Because of boom in ML and AI, dataset collection has become a business activity, and it can be done remotely. Something similar to Amazon Mechanical Turk but should not require heavy capital initially.