Landslide Detection using Deep learning Neural Network | Landslide4Sense | GeoDev

Поделиться
HTML-код
  • Опубликовано: 15 ноя 2024

Комментарии • 61

  • @Sandhya_Bytes
    @Sandhya_Bytes 10 месяцев назад +2

    Sir i dont know how to express my happiness for this wonderful content, This is really very helpful for my academic project. Thank u thank u sooooooo much. We need more content like this...........

    • @geodev
      @geodev  10 месяцев назад

      It is my pleasure! All the best and stay tuned for future similar tutorials

  • @MrSafeerUllah
    @MrSafeerUllah 2 года назад +4

    I really appreciate it, so far the best video about the ML application in natural hazards.

    • @geodev
      @geodev  2 года назад +2

      More contents related to ML/DL are coming on satellite imagery. Stay tuned!

  • @mario5554
    @mario5554 3 месяца назад +1

    Great Tutorial!!! Congratulations, and of caourse, thanks for not being jealous with your knowledge and thanks for sharing everything you did.

    • @geodev
      @geodev  3 месяца назад

      My pleasure! Glad you liked it!

  • @rahulds001
    @rahulds001 2 года назад +4

    Thank you so much...❤️ expecting more deep learning tutorial for geosciences application which is very rare in RUclips. No body teaches that

    • @rahulds001
      @rahulds001 2 года назад +1

      Can you tell me how can we merge this patches to create a single output if we have a geotiff image and how can we convert h5 into geotiff

    • @geodev
      @geodev  2 года назад +1

      More tutorials on the way. Stay tuned!

    • @geodev
      @geodev  2 года назад

      From this tiles, it is not possible because there are the tiles from random location. But for such work, I recently wrote the library named as "geotile" which will help you to create tiles as well as merge tiles. Library github: github.com/iamtekson/geotile

    • @rahulds001
      @rahulds001 2 года назад +1

      @@geodev Thank you so much. I would like to request one video on how to create our own dataset(patches and masks) from satellite data.

    • @geodev
      @geodev  2 года назад

      @@rahulds001 definitely, i will creat. Stay tuned😃

  • @yogeshkumardurugwar815
    @yogeshkumardurugwar815 2 года назад +3

    This video is very informative. Such types of tutorial are very rare. I am also working on landslide it will help me for my research. My best wishes for you and i am waiting for such videos on landslide inventory and prediction

    • @geodev
      @geodev  2 года назад

      Glad it was helpful! More videos are on way, stay tuned!

    • @yogeshkumardurugwar815
      @yogeshkumardurugwar815 2 года назад

      @@geodev sir please share with me your contact email id and ph. no. i have more work related to landslide, we can do with collaboration.

  • @kingLinger-m5d
    @kingLinger-m5d Год назад +1

    Great video. Kindly ask that for the Sentinel 2 images which platform was used to download it, maybe different imges download platform has diffenent outcomes. Thanks!

    • @geodev
      @geodev  Год назад

      I am also exactly not sure, which platform was used to download the original imagery. Please have a look to the published paper or landslide4sense website.

  • @Chaffy-v3p
    @Chaffy-v3p 9 месяцев назад

    the dataset provided, is it from a specific study area, im very new to ml and related topics but find it interesting. i wanna try this with a dataset used in a research paper for a specific geographic area, it state that it used Digital terrain model along with Enhanced Natural Terrain Landslide Inventory (ENTLI). it would also be beneficial for me to create something for a specific area. Also some papers mentioned other factors like rainfall and such, can this project use factors like that as a parameter. This topic was more complex than i anticipated so im asking alot of questions

  • @ibiswas8548
    @ibiswas8548 Год назад +1

    Very knowledgeable video
    I want to know
    What deep learning model is it ???
    Is it CNN??

    • @geodev
      @geodev  Год назад

      Yes it is CNN. To be more precise, It is Unet model

  • @nurainigama
    @nurainigama 2 месяца назад

    Do you have any references or literature to understand the theory of this method?

  • @roitai-dev
    @roitai-dev 10 месяцев назад +1

    Thank you so much professor.

    • @geodev
      @geodev  10 месяцев назад

      You are very welcome

  • @EcoresolveInc
    @EcoresolveInc 2 года назад +2

    Great video and important topic Tek!

    • @geodev
      @geodev  2 года назад +1

      Glad you think so-:) Thank you Mikey!

  • @shubhammaurya3822
    @shubhammaurya3822 2 года назад +1

    Great video

    • @geodev
      @geodev  2 года назад

      Glad you enjoyed it

  • @nemeziz_prime
    @nemeziz_prime 2 года назад +1

    Great video 👍🏻 hope you make more such tutorials

    • @geodev
      @geodev  2 года назад +1

      Thank you, Sure I will create more tutorials on deep learning. Stay tuned!

  • @davidpark2584
    @davidpark2584 Год назад

    Your lecture is very useful. Please let me know where I can download your dataset. I can't download it at this moment.

  • @niflag
    @niflag 7 месяцев назад

    Why would you set NoData to 0.000001 instead of 0 or 256?

  • @priyanthabandara1437
    @priyanthabandara1437 2 года назад +1

    Can I do similar analysis with LiDAR data? If it is possible please do a video please

    • @geodev
      @geodev  2 года назад +1

      I think it will be possible with RGB imagery along with LiDAR point cloud. At the end, we need to create the DEM/DSM for landslide detection.

  • @pavanchaganti1776
    @pavanchaganti1776 Год назад +1

    How to get mask data for validation dataset... its not provided by land4sense too!!
    Can u help with doing that

    • @geodev
      @geodev  Год назад

      For the validation set, you can test and generate the result. Sorry I haven't tested the model for validation set.

  • @OmenJap
    @OmenJap 2 года назад

    Did you only use landslide location to run the model? So, We don't need to include non-landslide points in the inventory data.

    • @geodev
      @geodev  2 года назад +1

      Sorry, I used the data from landslide4sense challenge, which is not geolocational data.

    • @OmenJap
      @OmenJap 2 года назад

      @@geodev okay, I understand that random forest method needs for dependent data such as yes/no or landslide/non landslide. But you run the model only with landslide data, is that right?

    • @rockyard100
      @rockyard100 2 года назад +2

      @@OmenJap Not really. In segmentation tasks you have masks/patches of the corresponding satellite image which acts as a label (for model training) and these masks/patches includes pixels from both the landslide and non-landslide class. So, such models can identify both landslides and non-landslides pixles and segment only the class of interest.

  • @sanjithhegde6008
    @sanjithhegde6008 4 месяца назад +1

    Thank you for the video sir 🙏
    Could you please help to know about How to import utils?
    Im getting error in importing

    • @geodev
      @geodev  4 месяца назад +1

      Hi, you need to write utils.py file as well. Please check the github repo and download the full code.

  • @shubhammaurya492
    @shubhammaurya492 Год назад

    Thank you Sir much awaited topic

    • @geodev
      @geodev  Год назад

      Always welcome

  • @LokeshSharma-jm4dp
    @LokeshSharma-jm4dp Год назад

    Can you please make such video for air pollution prediction😊

  • @neefiyasingh3232
    @neefiyasingh3232 Год назад

    I really like your videos. I want to know the dataset you have take above in landslide detection it's corrupted file. I tried to download train data and it shows file is corrupted. What should i do?

    • @geodev
      @geodev  Год назад

      Did you download the data from here: www.iarai.ac.at/landslide4sense/challenge/? I have worked on this data and it is not corrupted.

  • @VIGNESHMMsc2023VelTechChennai
    @VIGNESHMMsc2023VelTechChennai 2 месяца назад

    sir, i need full dataset link for project, thanks for the info

  • @zainabkhan2475
    @zainabkhan2475 7 месяцев назад +1

    bought your course on udemy

    • @geodev
      @geodev  7 месяцев назад

      Great! Thanks for the purchase.

  • @shilpakumari6874
    @shilpakumari6874 10 месяцев назад +1

    How can i use arcgis for data collection

    • @geodev
      @geodev  10 месяцев назад

      You can manually digitize the labels and produce the image tiles using "Export Training Dataset Using Deep Learning" tool.

  • @BinhNguyen-cp9dv
    @BinhNguyen-cp9dv 2 года назад +1

    How can I convert from TIFF to H5 or H5 to tiff, or any gis software can do it ?

    • @geodev
      @geodev  2 года назад

      H5 format doesn't come with coordinate system. But anyway if you want to export as a image, write it using rasterio or gdal.

  • @yuvrajthakur1821
    @yuvrajthakur1821 10 месяцев назад

    can you share your drive link where you have stored your project because i am not able to downlaod from site
    please

    • @geodev
      @geodev  10 месяцев назад +1

      Hi, I have removed the data from my drive but you can get the same dataset here: www.kaggle.com/datasets/tekbahadurkshetri/landslide4sense

  • @kaviyaakavi5969
    @kaviyaakavi5969 Год назад

    sir, Is this code is possible for real time images

    • @geodev
      @geodev  Год назад

      Yes, If you have an real time images, it will definitely works.

  • @slimshady6242
    @slimshady6242 8 месяцев назад

    arey hindi bol na

  • @Bithika_bera
    @Bithika_bera 6 месяцев назад

    I found your project, it's very well done, I have to contact with you for some problems plz help me

  • @jdkumar5588
    @jdkumar5588 2 месяца назад

    model = unet_model(128, 128, 6)
    #model.summary()
    checkpointer = tf.keras.callbacks.ModelCheckpoint("best_model.h5", monitor="val_f1_m", verbose=1, save_best_only=True, mode="max")
    #earlyStopping = tf.keras.callbacks.EarlyStopping(monitor='val_f1_m', patience=10, verbose=1, mode='max')
    callbacks = [
    #earlyStopping,
    checkpointer
    ]
    history = model.fit(x_train, y_train, batch_size=16,
    epochs=100,
    verbose = 2,
    validation_data=(x_valid, y_valid),
    callbacks=callbacks)
    model.save("model_save.h5")
    I am getting an error like this. What should I do to execute this code.

    ValueError: The filepath provided must end in `.keras` (Keras model format). Received: filepath=best_model.h5

  • @surihjagirani3336
    @surihjagirani3336 Месяц назад

    I really appreciate you, would you please share your email