Top Cat has a plan

Right, I uhh gots this plan See, for making an A.I so we's can identifies the artifacts see....

Okay, so, uhh, first things first, we gotta, uhh, gather a big bunch of pictures of, uhh, prehistoric creatures. Like, all the creatures you might find on, uhh, a prehistoric artifact, y'know? And, uhh, we gotta label 'em all so the computer knows what it's lookin' at. And, uhh, we wanna make sure we got a real diverse bunch of pictures, y'know? Different poses, different backgrounds, different lighting and all that jazz.

Next up, we gotta train up a, uhh, deep learnin' model to, uhh, recognize these creatures. Now, a convolutional neural network, or a CNN, that's a fancy way of sayin' a, uhh, computer that's real good at lookin' at pictures, is a good choice for this job. We feed it all the pictures we gathered earlier and, uhh, it starts learnin' what each creature looks like.

Once the model is train and validated on the test dataset, we move on to the fun part: makin' the app. Now we use, uhh, Java or Kotlin to make an Android app that can, uhh, take live video from the camera. Then we use the model we trained earlier to, uhh, scan the video frames in real-time and look for creatures. And when it sees one, we make sure the output is correct, and then we display the creature on the screen with its label and, uhh, save it.

Now, uhh, we gotta make sure everything is accurate and all that jazz, so we add a step where the user can, uhh, verify the results before we save it. That way we make sure we ain't missin' nothin'.

And once the user confirms it, we allow the user to categorize the artifact, and, uhh, send the data to a database or cloud storage. And, of course, we're constantly makin' improvements, y'know? Takin' feedback from users and, uhh, tryin' out new ideas.

But, uhh, just keep in mind, buddy, makin' an AI-powered app like this is a real big job. Takes a lot of expertise and, uhh, computin' power. But with the right team, you'll be the top cat in, uhh, prehistoric artifact recognition in no time!

Now in plain english:-The development of an AI-powered Android application for recognizing creatures in a live video feed of a potential prehistoric
artifact.

Collect a dataset of common creature depictions from prehistoric art. This dataset should include a variety of images of the creatures, along with labels indicating what the creatures are. The dataset should be sufficiently large and diverse to ensure that the model can generalize well to new images. Depending on the specific task, it might be beneficial to include images of the creatures in different poses, lighting conditions, and backgrounds.

Use this dataset to train a deep learning model for object recognition. A convolutional neural network (CNN) is a good choice for this task because it is designed to process image data and can learn to identify patterns and features in the images. A CNN typically has several layers, each of which extracts increasingly complex features from the input data. The final layers of the CNN are typically fully connected layers, which use these extracted features to make a prediction. Popular architectures like Resnet or VGGNet can be used. It is important to validate your model against a test dataset, to check it's accuracy, before moving to the next steps.

Develop the Android application using a programming language such as Java or Kotlin. The application should have the ability to capture live video from the device's camera and process the video frames in real-time. OpenCV library can be used for this purpose.

Use the trained model to analyze each video frame and identify any creatures that appear in the frame. The frames can be passed through the model, and the output should give you the predicted labels for the objects in the frame. The threshold for acceptance of the prediction should be adjusted to avoid false positive and negatives.

Display the identified creatures on the screen of the device, along with their labels, and also record the find. The user can save the image with the labeled creatures for later study.

Incorporate a method to validate the recognition before sending the data to be stored in a database. This could involve allowing the user to manually verify the results or using a secondary model to validate the predictions.

Allow user to categorize the artifact and submit the data to a database or cloud storage. This will help to keep track of the artifact and for easy sharing of the data.

Further optimization can be done based on the results and user feedback.

Some of the potential problems that may arise include:

Ensuring that the model is able to accurately recognize the creatures of interest even when they are partially occluded or appear in different poses or lighting conditions.
Managing the computational requirements of running the deep learning model on a mobile device. This may involve techniques such as model compression or quantization to reduce the number of parameters in the model.
Addressing any potential privacy concerns that may arise from capturing live video of a potential prehistoric artifact.
Making sure that the app can handle edge cases and various scenarios, such as poor lighting conditions, camera shake, or different camera angles.
It's important to note that this is a relatively complex task that would require a high level of expertise in multiple fields such as computer vision, deep learning, and mobile app development. Additionally, depending on the size of the dataset it could require a significant amount of computational power, which is can be expensive.

I had not thought of making an app, but have thought it would be great to have some sort of database to upload and compare similar figures/creatures that are commonly seen on rock art. I was going to set up albums in my Facebook group (Early North American Effigies and Rock Art) for this, but, unfortunately albums don't work the same way in groups as they do on my personal page. Or maybe I just haven't figured out how to do it. It is amazing how often I see other people post rock art that is nearly identical to some in my collection.

Yes, this is exactly how I realised this is world wide, recognising poly iconic artifacts from far diverse places containing the same icons, and the same conventions. The above idea is becoming increasingly possible, mobile phones are now being sold with A.I cores in there CPU's, rather than the model of uploading the images to a centralised server with enough compute power to analyse the video feed. The trouble with the current database is, or at least was when I last looked at a few years back is that it relies on the users to be:
A. Honest
B. Of significant skill to recognise genuine artefacts
C. the operator/owner of the database needs to weed out fakes and rubbish.

The beauty of A.I recognition is it takes out the subjective element the proponents of Portable rock art and figure stones are often accused of, Professionals cannot argue with a computer

Brett wrote:Yes, this is exactly how I realised this is world wide, recognising poly iconic artifacts from far diverse places containing the same icons, and the same conventions. The above idea is becoming increasingly possible, mobile phones are now being sold with A.I cores in there CPU's, rather than the model of uploading the images to a centralised server with enough compute power to analyse the video feed. The trouble with the current database is, or at least was when I last looked at a few years back is that it relies on the users to be:
A. Honest
B. Of significant skill to recognise genuine artefacts
C. the operator/owner of the database needs  to weed out fakes and rubbish.

The beauty of A.I recognition is it takes out the subjective element the proponents of Portable rock art and figure stones are often accused of, Professionals cannot argue with a computer

Some of the professionals I have debated with would argue with the rock itself if nobody else were around.