Definitely not with the dataset that we used. That would require training over a wider variety of images (weather conditions, vehicle types) and maybe even some augmentation to existing dataset
I referred to these 2: 1) medium.freecodecamp.org/tracking-the-millenium-falcon-with-tensorflow-c8c86419225e 2) github.com/bourdakos1/Custom-Object-Detection Instructions and github repo are very clear in the article. The repo isn't mine, I used his file structure as it was nicely maintained and there was a medium article with step by step instructions to go with it
Hi.. Mihir bhatia can you help me to select the trending techniques of research in aerial images.. plz provide your email too.. I will be very thank ful
Nope. I haven't updated it. But be sure to checkout and star the repo in the description (which I referred to for the project) . This person has updated the their documentation and code. Sentdex also has a TensorFlow object detection on a custom dataset video playlist
You can create your own dataset using images from Google Maps. Use ImageLbl software to automatically create a XML file with coordinates of the object detection boxes. For this project I used the article in description and Sentex's Tensorflow Object Detection API, be sure to check them out
@@majorfree3165 yes you wont acheive high f1 scores from maps images. Drone images would be great. Also, in my project I used just 50 random images. If you take the time to create a bigger dataset with 100s of images I'm sure you can get a much better performing model
Really helpful! Would like to know more about this project.
Thanks.
How can I detect vehicles using Sentinel-2 or Sentinel-1 satellite?
Did you consider how many vehicle for classification
Can you please explain further ?
@@mihirbhatia9658 For vehicle classification how many category of vehicles considered? (like car, bus...)
@@vinayarajvs8458 all vehicles are considered under one class
By which app u implemented this
Will this work if there cloudy weather or there is lot of snow on the car ?
Definitely not with the dataset that we used. That would require training over a wider variety of images (weather conditions, vehicle types) and maybe even some augmentation to existing dataset
I m working on object detection .similar to your project ,plz can I get your code
I referred to these 2:
1) medium.freecodecamp.org/tracking-the-millenium-falcon-with-tensorflow-c8c86419225e
2) github.com/bourdakos1/Custom-Object-Detection
Instructions and github repo are very clear in the article. The repo isn't mine, I used his file structure as it was nicely maintained and there was a medium article with step by step instructions to go with it
i have a project of predicting wheat using sataliite images. please help me with this. Paid project
These are airborne imagery, not satellite imagery, right???
We just used images from Google Maps and Apple maps to train. The test that we ran was on footage from a drone that we find found online
Hi.. Mihir bhatia can you help me to select the trending techniques of research in aerial images.. plz provide your email too.. I will be very thank ful
Not sure if I'd be able to help with selection of a topic. This was just my deep learning class project. Here's my email: bhatia35@purdue.edu
Can someone please share the code?
Hey did you do any update to your work ?
Nope. I haven't updated it. But be sure to checkout and star the repo in the description (which I referred to for the project) . This person has updated the their documentation and code.
Sentdex also has a TensorFlow object detection on a custom dataset video playlist
Hello, l'm working on the same project and I need a dataset
You can create your own dataset using images from Google Maps. Use ImageLbl software to automatically create a XML file with coordinates of the object detection boxes. For this project I used the article in description and Sentex's Tensorflow Object Detection API, be sure to check them out
@@mihirbhatia9658 thank you .I use 4k dron footage and now I'am extracting images from it. I tried map images but vehicules aren't clear
@@majorfree3165 yes you wont acheive high f1 scores from maps images. Drone images would be great. Also, in my project I used just 50 random images. If you take the time to create a bigger dataset with 100s of images I'm sure you can get a much better performing model
@@mihirbhatia9658 yes I'am working on it . can I get your Email.
@@majorfree3165 bhatia35@purdue.edu
@Mihir Bhatia