Why the images dataset is empty? Is there any source to access or download the image data. Plz provide the accessible link...
@dr.priyankarastogi5837
3 жыл бұрын
Hi Sir, In this classification, the train and test datasets have been split after resampling. Just a thought - is it a right way? Since same samples might be present in both test and train dataset especially in classes which have very few samples.
@MrJ3
Жыл бұрын
Agreed, this is a dangerous practice.
@digisportsonline9333
3 жыл бұрын
Sir please guide how we can extract features perform segmentation and train model
@lalitsingh5150
3 жыл бұрын
Sir please make a video on ELM classifier
@surflaweb
3 жыл бұрын
Great vídeo. Can I use data aumentation to balance my data? Thanks
@DigitalSreeni
3 жыл бұрын
Yes, data augmentation can be used to balance data but you also need to downscale majority classes. Augmenting 100 images to 1000 is not a good idea. I am comfortable with 2 fold augmentation but not 10 fold.
@tednones219
Жыл бұрын
do you have a dataset where all images are already separated into 7 classes in different folders?
@SwarnaliMollickA
2 жыл бұрын
Can I apply data augmentation to balance the dataset?
@DigitalSreeni
2 жыл бұрын
Sure. You can preferentially augment one class although you need to remember that Augmentation helps in generalizing the model but not necessarily improve the accuracy.
@SwarnaliMollickA
2 жыл бұрын
@@DigitalSreeni Thank you Sir!
@anmolgautam9572
3 жыл бұрын
This channel is such a great educational resource. Thank You Sreeni Sir for your efforts.
@DigitalSreeni
3 жыл бұрын
It's my pleasure
@ravinder_paspula
2 жыл бұрын
Brother can you do same operation with tf -encrypted..., I mean skin cancer classification on encrypted data set
@jasurmarufov7810
2 жыл бұрын
Really cool lesson and practice, thank you very much. I am doing a social project about this topic. If we get more data, not 500 pics for each one, but like 2500 or more, probably we'll get high accuracy?
@lucamaximilianesch2540
3 жыл бұрын
Where did you get all your knowledge about python? It‘s really incredible!
@haraldkittler4832
Жыл бұрын
Great video. Just one remark: The dataset is a courtesy of the Medical University of Vienna, Austria, not Harvard. Harvard is just hosting the dataset (but it can be obtained from other sites too).
@DigitalSreeni
Жыл бұрын
Thanks for the note.
@bhuvanashetty5853
2 жыл бұрын
Hi sir I'm beginner to python and image processing...Sir can we apply preprocessing and segmentation to ham10000 dataset for more accuracy ? If yes which techniques you are going to suggest ...
@locagagarin
Жыл бұрын
Hello, firstly thank you so much for your amazing videos.I would like to ask about model.fit when I try in the history = model.fit ( x_train,y_train, epochs=epochs, batch_size=batch_size, validation_data=(x_test,y_test), verbose=2) its gives me following error :'' Calling `Model.fit` in graph mode is not supported when the `Model` instance was constructed with eager mode enabled. Please construct your `Model` instance in graph mode or call `Model.fit` with eager mode enabled.'' can you give me advice
@ravinder_paspula
2 жыл бұрын
I need your help to implement breast cancer dataset classification with tf-encrypted ..please I am ready to pay as per your interest...
@godomotives7171
3 жыл бұрын
This is perfect sir, you are a great lecturer. I am requesting for your guidance on the steps I can follow in order to learn coding and programming within 3 to 6 months. Please put them in order from first thing I should do to the last. Thank you.
@prayassoni7876
2 жыл бұрын
Hii at 11:38 you said that the images will copy itself using resample () from sklearn. I have a doubt regarding that. Does the images as in form of jpg's get copied and are upscaled to 500 or its in the form of numpy array which gets upscaled.? Thanks
@swatimishra1555
4 ай бұрын
How can I use augmentation in this? Please tell
@swatimishra1555
4 ай бұрын
if anyone can help, plz tell
@MalikMehsi
6 ай бұрын
Came for the explanation, stayed for the Packers Helmet in the back! Go pack go!
@Dr.Celine-xb8bh
Жыл бұрын
excellent video, where I can find the HAM image base that all images are already separated into 7 classes in different folders.
@gtxbiscuit799
3 жыл бұрын
When I try to apply random forest, there shows an error while fitting the model that the x_train has 4 dimensions, while we can only fit the model with two, can you guide me on how to convert the 4 dim to 2?
@guntikarthik7905
11 ай бұрын
Hlo... sir Can we apply random forest classifier after train test split?
@yemisijaiyeoba2652
Жыл бұрын
Thank for this video Plz what RAM size is best to run this code?
@Camila-fv9qj
Жыл бұрын
Why you used confusion matrix not f1-score or auc- roc ..etc
@chronobreaker2374
Жыл бұрын
Sir what are system specification as i want to buy a new system
@javariatahir9405
2 жыл бұрын
Great work, sir can you please make a tutorial on transformer for segmentation (TransBTS paper).
@alejandroalonso2002
9 ай бұрын
best channel on KZitem thank you!
@dr.aravindacvnmamit3770
3 жыл бұрын
great work sir ????? I need one thing its request can we identify exactly the area and its stages automatically is there any techniques. I tried with Grad-Cam sir can you show if other ?
@DigitalSreeni
3 жыл бұрын
If you want to identify these features in a large image you need to look into mask R-CNN approaches.
@surflaweb
3 жыл бұрын
Hi Sreenivas please don't forget to upload the code to github..
@DigitalSreeni
3 жыл бұрын
There is always a bit of lag between video upload and github push. Sometimes I forget and viewers remind me :)
@surflaweb
3 жыл бұрын
@@DigitalSreeni Thanks man
@aysenur3057
2 жыл бұрын
Hello, I am just curious why didn't you split the dataset into 3 part? If I am not wrong, you splitted the data as train and test and used the training part in training phase and the test part while training phase as validation data. Then you tested your model with that same data. According to my knowlwdge, validation data provides syncronized validations and updatings to parameters while training continues. And test data is used for measuring real accuracy thus that test data should be hasn't seen by the model before. Did I miss anything, please explain.Thanks for the content, lovely lecture.
@DigitalSreeni
2 жыл бұрын
The test data does not get used to update the weights (to train) during the deep learning training process. It gets used only to assess the model quality using the defined metrics. The training happens only using the training data. Therefore, you can use the test data also as validation data. In real life data sets, you split your data into train and validation. The validation gets used to report the metrics during and after training. You then obtain a separate test dataset (NOT part of the whole training data) to really assess the quality of your model. Just splitting your large data set into train / validation / test is not going to tell you anything more than your train / test split which is why I always split my data into two sets. The goals for training a model is to use it in real life. This is why your test data always must come from outside of your training data. Normally, I collect new images from a microscope or MRI or whichever method you use to collect images. In this specific example, you can just google search for some skin cancer images and see how the model performs.
@aysenur3057
2 жыл бұрын
@@DigitalSreeni thank you a lot for your super clear explanation. I don't know how can I thank you. it helped a lot. please let me ask another question which is about data types. when I get my image column and convert to list and numpy array, I see my X, x_train, x_test are numpy array but keeps string. I think because of that when I give it to model, I ge an input shape error like "Model was constructed with shape (None, 224, 224, 3) but it was called on an input with incompatible shape (None,)." thanks a lot 🙏🙏.
@ravipuvvula6452
2 жыл бұрын
I used the main dataframe and split it into train and valid first. Then split the valid into two parts to get all train, test and valid dataframes. Hoping to get 95% ACCURACY 😅
@kukrak
3 жыл бұрын
Hi, how to predict class of one concrete image from file?
@DigitalSreeni
3 жыл бұрын
Load the image, pre-process it exactly the way you processed images for training, then do model.predict.
@vimalshrivastava6586
2 жыл бұрын
Thanks for all your efforts 🙏
@kumudpatel9124
3 жыл бұрын
sir, can you please make a video on time series project
@DigitalSreeni
3 жыл бұрын
I have a few videos on my channel on time series.
@sabaal-jalal3710
3 жыл бұрын
Thanks Sir , have learnt a lot from your channel.
@DigitalSreeni
3 жыл бұрын
Glad to hear that
@VemulapalliVinayKumarBEC
Жыл бұрын
how can i open github code?
@DigitalSreeni
Жыл бұрын
Download (or copy paste) to your favorite IDE. I often use Spyder that comes with Anaconda installation.
@nafisatasneem6198
2 жыл бұрын
This is a great tutorial for beginners! However, while trying to add the resized images in the dataframe, by using 'skin_df_balanced['image'] = skin_df_balanced['path'].map(lambda x: np.asarray(Image.open(x).resize((SIZE,SIZE))))' code, it is showing error. The error says, "'NoneType' object has no attribute 'seek'". Can you guide me about how can I resolve this issue? Thanks in advance!
@DigitalSreeni
2 жыл бұрын
Whenever you see "NoneType" in the error message, it means there is no data for your method to work. Please make sure your images are actually captured as numpy array and that the arrays are not empty. Run one line at a time to figure out the issue.
@nafisatasneem6198
2 жыл бұрын
I have resolved the issue. Actually, I uploaded the sub-folder dataset into google drive and was working in google colab. I found that there were some missing images in one of the uploaded subfolders (nv). For this reason, the error was showing. Thank you for your feedback.
@spam.moizah
2 жыл бұрын
@@nafisatasneem6198 i am facing the same issue can you help me
@nafisatasneem6198
2 жыл бұрын
@@spam.moizah are you working in google colab? Or your local machine? If working in colab, have you uploaded the dataset in google drive? If so, chances are high that the uploaded dataset is incomplete (this may happen as the HAM10000 is a large dataset). Try to check if the uploaded dataset has all images.
@spam.moizah
2 жыл бұрын
@@nafisatasneem6198 can you tell me your Gmail account I will contact you on that if you don't mind
Пікірлер: 68