如果不是真正理解並專精遙測的理論/原理與技術/方法 我真的很懷疑如何整合並利用越來越多的各式各樣遙測平台所提供的影像資源進行DATA MINING與DEEP LEARNING?
當然 這前提是已經具備DATA MINING與DEEP LEARNING的理論基礎與技術.
I am not saying that you can throw away GIS --
Actually you need to know more in depth how to integrate GIS and RS so that you can swim freely in the ocean of huge data as a fish of data scientist.
Once, someone told me that he would teach vector GIS and I would be responsible for raster GIS. What were you talking about? you must be out of your mind........
How could you seperate vector from raster during data processing and analysis.
"我一直很好奇，商業系統中如何整合geospatial data，尤其是在deep learning中談到Multiple-Model Machine Learning .....，他們花很多年時間建立許多pipelines
我現在工作跟以前沒兩樣，只是不用ESRI products, Erdas Imagine, PCI etc. 但一直覺得不夠有效率，應該有統一的pipelines，去整合不同的projects
"GREAT (not just good) question.
You hit the key point that why I prefer ESRI coz they have big bosses keep pushing them for better integration -- though still a long way to go, but better than others.
For people who just want to play with, bush around, open sources and R/PyThon might serve their needs "OK". For serious jobs, I'll bet money on big names with complete system design for integration (including what you just mentioned pipeline ....)
I know people who are working on super computers / parallel processing have their own protocal and very specific/unique algorithms for pipelining.
Maybe it's time for you to investigate this domain!
As to the issue of pipelining between vector and raster, it's a mission impossible to integrate vector and raster as a symphony. The only way is to invent sort of magic that could convert between these two totally different monsters back and forth on the fly. Yes, we need Merlin the wizard....."