FAQ

How It could identify objects.

Unlike other apps that use crowdsources tagging services, that involve people that manually add tags to pictures. The detection in Image Analysis Toolset is entirely driven by deep learning for computer vision, this means that only advanced artificial neural networks handle the loaded pictures without manual human interventions.  The system behind IAT uses artificial neural networks to mimic processes involved in the human vision using the concepts discovered by neurobiology on the way human brain work, that has proved how several types of neurons were activated only under certain stimulation these concepts have been applied to deep learning for computer vision techniques to create models that become more and more accurate.


I get a message about adblocker but I haven't any adblocker.

Probably you are using some modded ROM that includes an adblocker function at host level. 

To use the App in free mode, you have to remove admob and googleads related lines from host file located in /system/etc/hosts editing it with a text editor (by default this file should be blank since contains the list of address to block). You can use a host editor like this to simplify the procedure. Remember that all host file changes usually are applied after device restart. 

Ads are needed to maintain the free service, if you purchase a premium subscription all adblocker checks are disabled, however please note that App doesn't care if you have browser-only adblocker or whatever other kind of adblocker that doesn't interfere with the ads of our App.


I have uninstalled Adblocker but the access is still denied. 

Some adblockers apply a patch to the system host file and uninstalling them doesn't change anything. Open the adblocker and check if It has a command like "disable adblock" or "restore original host file", then restart and try again. If there isn't such command in the adblocker you use, try to follow the instructions in the answer of the previous question.   


I have whitelisted IAT in my Adblocker but the access is still denied.

For some adblockers the changes are applied only after the restart of the App and\or device. Try to restart the device. 


Why do I see the message about the choose of the path for default App folder every time I start the App?

You haven't selected correctly the default App folder path, see the next FAQ question for instructions about the way you should do it.


How to select the default App folder path?

When the folder chooser is opened, it may show recent opened folders section or others paths not selectable. If the path isn't selectable the button to select the path is missing or greyed out and you have to move in a path writable (e.g., Documents) or create a new folder and then select it. In case you cannot see your internal storage, click on the top right menu of the folder chooser and click on "Show internal storage", now you can access the internal storage, opening the drawer menu on the left.


Where are the premium license options?

You can access the premium license options clicking the key icon that appears in the top bar of the home section.


Commands in the home of the App don't work, app stuck or infinite loading.

Google Play services are updated by the OS automatically and sometimes something could go wrong and since Google Play manages also the App licenses, this could cause issue to the App. If the App commands don't respond probably your Google Play cache is corrupted. 

In most cases is enough to restart the devices to fix Google Play cache issues, if this doesn't happen you have to fix the issue cleaning Google Play and Google Play Services cache following this procedure:


There is a message related to HIGH_SAMPLING_RATE_SENSORS permission in the log.

This permission is needed only on newer versions of Android (version 12 and newer) for accurate movement correction during real time analysis, before, apps could use high sampling rate sensors, without the need of this specific permission.


When IAT will be translated in my language?

The localization of IAT is a complex matter. The App interface itself could be translated like the others Apps, but the results you get are a different matter. The numbers of element that IAT can recognize is huge and grows continuously, so we are working to a system to translate the element results automatically without any risk of translation errors that could be introduced by classic translators. In settings you can enable the option "detector translation" to translate the elements in your language but the system that we have in plan isn't ready, so this feature could introduce some translation error at the moment. If you want to speed up the localization progress you contribute to the App interface translation follow instruction at this link or contact us.


App hasn't recognized something.

You can try to increase the quality of detection in settings and\or try to do a clearer picture from a frontal angulation. If this doesn't solve, please note that We haven't access to the images that users analyze so we cannot help you if you don't try to send us a mail with the image.


I would prefer a faster but less accurate detection.

Although the 1200 value has a good trade off accuracy\speed, in settings you can try to decrease quality to get faster results.


I would prefer a slower but more accurate detection.

Although the 1200 value has a good trade off accuracy\speed, in settings you can try to increase quality to get better results.


Censorship risk meter reports me as at risk an image that shouldn't be problematic. 

This feature hasn't been introduced with the purpose to be as accurate as a human in judging at risk images, but to report risks related to the images use in the web, especially false positives that you normally wouldn't guess as problematic. Often many social users get random images deletion (especially artists and fashion models) without understand the reason, this due to automatic systems that applies automatic removal because of false positives. Some years ago, a famous Facebook Page about scientific research has posted the picture of the mole rat (Heterocephalus glaber) that you see in one of the screenshots and It has been banned for many days: cause some automatic system has wrongly interpreted It as pornographic content. With this function you will be able to see if a picture that respect all the rules has been banned for no reason due to some error of auto-analysis used by the various websites, and you could also check if a "borderline picture" could trigger some censorship measure. 


Why EXIF info function shows inconsistent information?

EXIF metadata are bonded to the image file structure and some websites reprocess the images creating a new image file where the original EXIF metadata are totally omitted and the RGB matrix reprocessed. In case you have tried to analyze the image's EXIF using the share image function, the lack of EXIF information could be also caused by the App from where you have shared the image, because some Apps share the image preview rather than whole original image file. In this case you may try to open the image file directly from IAT to see the complete EXIF information.


Why some features don't work offline? 

We are working to improve the offline mode. Some features like element detection, face detection and OCR have already an offline mode, others like barcode detector and EXIF analyzer works fully offline. For element detection however this cannot be anyway near the accuracy of the server-based element detection. Mobile devices aren't powerful enough to run the software behind cloud-based element detection, and would be needed also a lot of storage to contain the huge database used.  We will try to make working offline all features possible in the best way possible, it's also in our interest, since features working offline can reduce our Cloud server load, lowering our costs. 


My connection is active but IAT continues to tell me that I need a connection active. 

This problem can have various causes. Very rarely It could be a service down during maintenance tasks, in this case the issue is temporary and very soon the App will work again, but in most cases It's due to the fact you haven't installed the App from Google Play and you are using an unregular banned version of the App that cannot access cloud server anymore (or a very old App version that cannot access the server due to change in the access interface). You should download IAT only from Google Play. We distribute IAT officially only on Google Play, if you download the apk of IAT elsewhere you can get unsafe modified Apps that can include malware and spyware and no antivirus in the world can fully protect you, since they can only detect old and already known threats. If you use a VPN another possible cause is that VPN node that you are using has been involved in malicious activity and cloud refuses to respond. 


What are the Custom modes in Real Time Detection and Element Detection?

IAT offers an advanced feature to allow power users and computer vision researchers to load analysis models trained by themselves. The models must be TensorFlow Lite models. The TensorFlow Lite models (with extension .tflite) must be placed in "Image Analysis Toolset\TFModels" path. When the files have been placed It's enough to configure in the setting menu the model name to use for the custom detection features.  Custom models could be used for identification tasks [I] or localization tasks [L]. The custom detection is available also in batch analysis, if no model is provided will be used the default model. IAT Custom Model Maker, an easy-to-use model creator software, has been developed in order to simplify the process and allow everyone to create his own custom models.


Where can I download IAT Custom Model Maker?

A preliminary version of IAT Custom Model Maker could be downloaded from HERE (win) or HERE(linux). At the moment the software is an alpha version available for Windows x64 and Linux x64, in future it will be released also for Mac and different architectures. This tool makes the creation model process very easy and accessible even for people with no experience with computer vision deep learning and related software stack configurations, anyway note that if you are a field expert you can use also others tools to create the models, if you respect the supported configuration parameters detailed in "How to configure Custom TensorFlow Lite Model" section.


Could I use as custom models third-party models not created with Custom Model Maker?

If models are in a supported format (as specified in the next FAQ) they can be used with the app regardless the tools used to train them. 

Here some sample models used for research purposes that could be used with the app: 

The downloaded model files can be renamed (maintaining the extension) with names more practical to show in the App menu to identify them, since model selection name depends on file name. Note that not all models can be used for localization mode [L], most third-party models are classification models suitable only for identification mode [I]. Consider also that since custom models runs totally locally on your device, if model is too large your device could be not powerful enough to run it.


How to configure Custom TensorFlow Lite Model?

The custom models must have the extension .tflite and have to follow these specifications:


The official App link. 

Latest version: 0.3.3 SR 190Apk Hash SHA-1: 58FCFC1813B73644CA52DD7B30028CB24C65E027