I want to create custom dataset for yolo9000. I read few articles on the internet but I could not find any clear solution mentioning the input format.
In general yolo accepts [id, x, y, x, y] as an input format for each object. In the similar way I would like to know the format for yolo9000.
I found this on github, but I am unable to understand completely how the data might have been created. What does n00002452 indicates here? Is it the image?
Can anyone share more information on this, or help me understand what does the input looks like for a single image for yolo9000?
Related
I opened a thread yesterday asking how I should be encoding dates and times. It was suggested that the datetime object would be a good format, as in 'have the least issues'. Here's an example where date time doesn't seem to be working.
Can anyone help me understand why? I've noted an increasing tendency of Vega-Lite examples to use ISO-8601, and in fact if I change this example to use that format it works. So this is more of a 'why isn't it working with datetime'. Is this format still recommended? Does it still even work?
Apologies but I think I misled you with my comment on the other thread (now deleted). Date time objects are only used for filter transform, scale domain, and axis/legend values. I have looked through all the sample json files for Vega Lite and can find no example of specifying a data object like you do in your sample.
Instead, if you specify you data as follows, you don't need to provide a format object.
{"entity": "Wildfire", "date": "2022-10-01", "deaths": 75}
I've been working a bit with some files from Minecraft Dungeons, which were extracted using QuickBMS and made available here: https://minecraft.fandom.com/wiki/Minecraft_Wiki:Minecraft_Dungeons_game_files
In the "data" folder, there are a bunch of json-files, which I believe contain a list of textures associated with any given stage of the game. There is, however, a problem. When opened, it reads like any json-file, it has a bunch of names and values, but some of the values are not human-readable, they instead show up as a string of seemingly unrelated characters. Here an example:
"walkable-plane" : "eNpjYSEOMIMAOp+ZmQmND1fEjF2AiQldAJsWDEPRXUKkowkDAM/qA6o=",
Now, given that these are exclusively characters, and not error signs or something of the sorts, I'm assuming this is an encoding issue. Of course, I don't know for sure, Or I wouldn't be asking this in the first place, but the file as it appears in the text is UTF-8, and it obviously doesn't produce a usable result. So, if anyone knows what exactly this is, and how I could extract information from it, I'd be really thankful.
I am using nodejs.
I am having the following problem -
There is a very big json object which i want to map to a much smaller object with a specific format which i will be using later on.
I want to have specific schemas which i will be able to use on that big object, each schema will have its own purpose. I want to be able to customize the structure via those schemas and no where else.
I was searching for a library which can help me with that.
I found one via npm which is called deep-map.
https://www.npmjs.com/package/deep-map
I played with it for a bit and it seem to answer my basic needs.
But i also going to be needing to do some more complex mapping.
A simplified example -
"testObj": { "myArr": [ {"type": "x", name:"test1"}, {"type": "y", name:"test2"}] }
and i need to look in myArr only for the name for which type equals x.
So basically i need some sort of for each loop and if term in between.
Since deep-map uses lodash/template i thought maybe there is a way to use the capabilities of lodash to solve this problem but so far i didnt find how to combine those two ( lodash + deep-map ) to solve the more complex mapping.
I am also open to other libraries which might help me with this problem.
Lodash has an at method that will do the simple cases of this. It lets you specify things like field.subfieldThatsAnArray[1].subsubfield and get the value.
For less of a "built in your own" solution, there's a couple bits of tech that might be interesting here too:
JSONPath
SelectTransform
There's some others out there, but JSONPath is used some other places in the space, and SelectTransform just looks cool.
I've looked at the very similarly titled post (Is there a C equivalent to Perls' Dumper() method in Data::Dumper?), regarding a C equivalent to Data::Dumper::Dumper();. I have a similar question for the Go language.
I'm a Perl Zealot by trade, and am a progamming hobbyist, and make use of Data::Dumper and similar offspring literally hundreds of times a day. I've taken up learning Go, because it looks like a fun and interesting language, something that will get me out of the Perl rut I'm in, while opening my eyes to new ways of doing stuffz... One of the things I really want is something like:
fmt.Println(dump.Dumper(decoded_json))
to see the resulting data structure, like Data::Dumper would turn the JSON into an Array of Hashes. Seeing this in Go, will help me to understand how to construct and work with the data. Something like this would be considered a major lightbulb moment in my learning of Go.
Contrary to the statements made in the C counterpart post, I believe we can write this, and since I'll be passing Dumper to Println, after compilation what ever JSON string or XML page I pass in and decode. I should be able to see the result of the decoding, in a Dumper like state... So, does any more know of anything like this that exists? or maybe some pointers to getting something like this done?
Hi and welcome to go I'm former perl hacker myself.
As to your question the encoding/json package is probably the closest you will find to a go data pretty printer. I'm not sure you really need it though. One of the reasons Data::Dumper was awesome in perl is because many times you really didn't know the structure of the data you were consuming without visually inspecting it. With go though everything is a specific type and every specific type has a specific structure. If you want to know what the data will look like then you probably just need to look at it's definition.
Some other tools you should look at include:
fmt.Println("%#v", data) will print the data in go-syntax form.
fmt.Println("%T", data) will print the data's type in go-syntax
form.
More fmt format string options are documented here: http://golang.org/pkg/fmt/
I found a couple packages to help visualize data in Go.
My personal favourite - https://github.com/davecgh/go-spew
There's also - https://github.com/tonnerre/golang-pretty
I'm not familiar with Perl and Dumper, but from what I understand of your post and the related C post (and the very name of the function!), it outputs the content of the data structure.
You can do this using the %v verb of the fmt package. I assume your JSON data is decoded into a struct or a map. Using fmt.Printf("%v", json_obj) will output the values, while %+v will add field names (for a struct - no difference if its a map, %v will output both keys and values), and %#v will output type information too.
i would really appreciate any help.
I would like to convert this list
[[{id1,1},{id2,2},{id3,3},{id4,4}],[{id1,5},{id2,6},{id3,7},{id4,8}],[...]]
to a JSON object.
Need some inspiration :)
please help.
Thank you.
Since you asked for inspiration, I can immagine two directions you can take
You can write code to hand-role your own JSON which, if your need is modest enough, can be a very light-weight and appropriate solution. It would be pretty simple Erlang to take that one data-structure and convert it to the JSON.
"[[{\"id1\":1},{\"id2\":2},{\"id3\":3},{\"id4\":4}],[{\"id1\":5},{\"id2\":6} {\"id3\":7},{\"id4\":8}]]"
You can produce a data-structure that mochiweb's mochijson:encode/1 and decode/1 can handle. I took your list and hand coded it to JSON, getting:
X = "[[{\"id1\":1},{\"id2\":2},{\"id3\":3},{\"id4\":4}],[{\"id1\":5},{\"id2\":6},{\"id3\":7},{\"id4\":8}]]".
then I used mochison:decode(X) to see what structure mochiweb uses to represent JSON (too lazy to look at the documentation).
Y = mochijson:decode(X).
{array,[{array,[{struct,[{"id1",1}]},
{struct,[{"id2",2}]},
{struct,[{"id3",3}]},
{struct,[{"id4",4}]}]},
{array,[{struct,[{"id1",5}]},
{struct,[{"id2",6}]},
{struct,[{"id3",7}]},
{struct,[{"id4",8}]}]}]}
So, if you can create this slightly more elaborate data structure then the one you are using, then you can get the JSON by using mochijson:encode/1. Here is an example imbeddied in an io:format statement so that it prints it as a string -- often you would use the io_lib:format/X depending on your application.
io:format("~s~n",[mochijson:encode(Y)]).
[[{"id1":1},{"id2":2},{"id3":3},{"id4":4}],[{"id1":5},{"id2":6},{"id3":7},{"id4":8}]]