Official Twitter Account
Get Support
Create Support Ticket
Share Page
Partner List

Simple yet very apreciated update for Image Recognition action

ReplyThanks 2017/05/01 00:11:54 0 0

I have been working on a bot that automates the process of manual traffic exchange on a well known site. What I am missing, I was under the assumption that this is already in, is the ability to supply an input image file link (from the disc, reference image is already saved locally) to the Image Recognition action. Another alternative would be to update the reference image every time the command is run (by specifying some coordinates from which the image should be cropped).

The current approach is that the reference image is saved only once (when the action is configured). Unfortunately this will not allow for more advanced operations to be done.

P.S.: I have plenty of neat ideas of stuff to automate and I want to share some of the best free-of-charge with the world. How should I proceed with this? If I post it on the forum and people like the module/projects, is there a chance they will be included in the runner?

2017/05/02 15:19:41

Did you want to input the photo link on the photo setting place?

Using the unknown photo probably will cause it cannot recognize it successfully. So this operation will be meaningless.

2017/05/04 00:00:16

not necessarily a photo link (I can also provide a image file from disk by specifying a file path) . One example for my usecase. There is at least one traffic exchange surf site which offers a reference image that you have to find and point to within 4 slightly different images (one of them is the same with the reference image). I can then use the bot to save the reference image to disk and then the image recognition to match it to what I have to click on to proceed with the next site in the traffic exchange service.

I don't see why it wouldn't work as you suggest (please take in mind that the image recognition block already allows us to specify a match degree).

2017/05/04 11:15:13

Here are two reasons caused it cannot work:

1. this action -image recognition can only recognize the exactly same image. If the image's size, color and pixel is different from the website, then it cannot recognize it or it will return a error message.
2. the image in the website must be in the running window’s visible range, then this action can recognize it. If you need to scroll down your mouse to see the image, then this action cannot recognize it.

2017/05/05 02:34:36

Below is a screenshot of how it looks. Basically I either need to tell the Image Recognition action to always cut at a fixed position (to crop the first violet car), or to supply it as a parameter the path to the image of the violet car saved to disk (this part I have already automatized with BotChief).

Then I need the Image Recognition action to match the second violet car. At every different surf I get the image of the first car in a different color. Its position in the last four car list will vary. This is just one example (there are a lot of sites resorting to such an approach to deffer against bots). As you can see in the screenshot below this is far different from what a image captcha would involve. This is why I think solving this issue with Image Recognition would be better suited.

P.S.: Thanks for the interest so far.

2017/05/05 11:12:38
If there are multiple same image on the surf. It will only go to match the first one which is exactly the same.

<< < 1> >>
Advanced Option