I have been looking for ways to use specific body landmark keypoints such as only the left and right shoulders from the keypoints as designed by the COCO keypoints 2017 dataset for human 2D pose estimation. However, I am unable to find consistent information anywhere on what is the order in which landmarks are arranged. Specifically, I am looking forward to map each (x,y,v) for each landmark to each body part. They change depending on which website I refer to. Can anyone please point me to the right reliable resource for it?
Thanks.
Related
I have been creating a twitter bot to post satellite images of random coordinates and I have found that the vast majority are of the ocean, so I would like to somehow make the bot detect that the entire photo is blue and repeat the process. The bot code is available here.
Thanks in advance.
As an alternative to employing machine vision to assess the content of the image to determine whether to sample again, you could simply sample from coordinates that are not the ocean! There are many files produced by geographers that have sought to map the boundaries of the world's continents.
For example, see here for code to do so. You'll need to understand projections and other features of Geographic Information Systems if you want to get extremely accurate, but you can get reasonably accurate using the polygons for oceans. If you REALLY wanted to get accurate, you could obtain a set of shapefiles that partition all possible coordinates (rather than a single "world" shapefile) and use the more-detailed mapping that may accompany those more-zoomed-in shapefiles.
Alternatively, you can implement the model here, provided in a guide to water detection in satellite imagery using Python.
I am trying to retrieve information about trees surrounding a given location from the Socrata API.
API Endpoint Description
I found two functions within_circle(...) and distance_in_meters(...) which I could use to filter the data set. The problem is, that those functions need either a location or a point data type which is not present in the data set.
There is, however a latitude and longitude field.
Is there any method to utilize those functions or get nearby trees other way?
Tried this, but POINT(0 0) must be the point of the tree.
https://data.cityofnewyork.us/resource/nwxe-4ae8.json?$where=within_circle( 'POINT(0 0)' ,0,0,400)
I need something like this.
https://data.cityofnewyork.us/resource/nwxe-4ae8.json?$where=within_circle( make_point(latitude, longitude) ,0,0,400)
SoQL
If you have not already done so, you may want to submit this question at https://support.socrata.com. In addition to the possibility that the people there will have an answer I do not, it would serve as some feedback that a function like what you had in mind would be useful.
I cannot think of a way to do exactly what you have in mind. Really, what I mean is a way that is within your power (or mine). The owner of the dataset could create a Point column -- and you may want to reach out to the NYC open data team to ask for that if you have not already done so.
However, since the X and Y coordinates, in feet, are present, you should be able to use the Pythagorean theorem to determine the distance from any given point. For that matter, the size of a degree of latitude or longitude cannot vary that much over an area as small as NYC so you could do the same thing with those values and save having to figure out the X and Y of your reference point.
Good luck!
Jon
I want to measure the distance to the next intersection/traffic light based on a gps location(s).
I DON’T have a route (i.e. no destination point) but I can get gps locations every second (while on the move).
Is it possible to get the information using HERE APIs?
I understand that there is no direct way (specific api) to do it, but I would appreciate any idea that will fulfill my needs.
The general idea that I came up with was getting at least 2 gps points to understand the direction, then getting link_id for current location and try to get the next links in same direction/road and find if any of them are intersections...
I didn’t find a way to do it. Any ideas?
There is no direct API for this however there is an idea that you have to crawl along the link into the direction determined like you outlined (or better by sending the last few GPS points to our Route Matcher (aka RME)) until you reach an intersection.
And this can be achieved with a concept of "electronic horizon" in SDK. It takes your current road link and driving direction and then crawls forward over the upcoming links until an intersection comes (or even beyond taking the most probable choice). mobile SDK has this "electronic horizon" feature.
https://developer.here.com/documentation/android-premium/dev_guide/topics/electronic-horizon.html
Im developing 2d isometric game. You driving car on city, thats all.
The issue is generating random maps with connected roads.
I would like to write script - i think i will be able to - but i cant find an idea for alghoritm itself.
Lets say i have 100x100 map, and i would like to build boolean table 1 is road 0 is not road.
As far i have solution that is drawing random number of lines (4-8 for ex) in horizontal and same vertical. But this road map is straight.
Can u share some ideas? Any will help
the question is too broad for a simple answer. theres plenty of ways in various degrees of complexity.
but as tip, consider the problem from a different perspective. sure in the end you want roads, but why are the roads there in the first place?
in your case its a city. a city consists of city blocks wich usually are rectangular. so one way would be to find a way to fill the map with rectangular shapes and consider the edges roads (or only some).
or you could look into triangulation algorithms and triangulate your map with a bunch of random points. then combine some triangles and use those edges as roads.
or even only use a random walk set up so it doesnt turn around completely.
or ...
... seriously though, the options and possible solutions are manifold and dependent on your skill level (as well as how you want to look your city in the end).
dont search for your specific problem, try to adapt some other algorithm for your need. theres plenty of tutorials on random dungeon generation for roguelike games.
also in the end there probably wont be a single generation algorithm giving you the best result, but a combination of many.
Can anybody give me any idea about what kind of traffic / sample size I need to get a statistically significant result when doing a google content experiement for 2 variations?
Google uses Multi Armed Bandit testing.
Here is a good article on this Googles answer
The best way in practice is to watch the percentage in the Google analytics experiments tab and see how quickly it moves toward 95%.
You can't get an exact answer because it changes as you take measurements and based on the difference you are trying to measure. So if one variation performs 300% better than the other it will take a lot smaller sample size than if one variation only performs 10% better than the other.
To see how the math for straight up statistical significance works here is a good explanation. Statistical significance tutorial
Here is a spot where it has a calculator Calculator
As far as the math for the Multi Armed Bandit this quote by Peter Whittle sums it up
[The bandit problem] was formulated during the [second world] war, and efforts to solve it so sapped the energies and minds of Allied analysts that the suggestion was made that the problem be dropped over Germany, as the ultimate instrument of intellectual sabotage.