Week 01

October 03

Thesis outline. Paste the results to the outline as we go along
Send email Jerry Weimer about if I can access the code? for NASSCO PACP.
Send email to Pipelogix to see what is the difference between pipelogix and Phoenix
Send email to Sacramento Area asking if they follow PACP standards
Read Joakin's people why he uses Gaussian blur to block the text.
Send instruction to Pavan on how to create CSV file from json (fname, label, distance, dates, clock angle)
email to author that is on the leader board. Asking him about the model
email Joakim how he did gaussian blur
send email to deven and sudhir the paper of segmentation and water level
Sudhir thinking of leverage the flood segment techique to find the water level in the sewer
If they inspect after the storm, we can we use this to estimate the I&I

Run the model without algebraic loop and save the results
Run the model without algebraic loop with interpolation = 0.002 and save the results
Check if the results are the same
Run the model with delay block and save the result and check if the results are the same
implement the delay block to the model with the algebraic loop to see if it resolve the problem. (no)
Implement the hyddelay block to the model resolve the problem with the algebraic loop.

Collimator: the flow into the Activated Sludge is lower than Matlab's value by 10,000 - 20,000 m^3/day. I will check my flow combiners block. But first, I will replace the submodel block with a single python script and then debug from there.
EPANET: After increasing the time step from 5 min to 1 hour, I got the system flow graph when running PDA on unfix pump file. See the graph in the attached PowerPoint.
CCTV: You need to include a transform when creating dataloaders to ensure all data of type TensorImage is the same size
470 X 700 is the standardized DNV image size (from Pavan)

Email: use accelerate seems to have GPU fragmented (3288). We will share your comment with accelerate
Run the EPANET with net 6 example.
CCTV: Add the metric for code. Figure out how to add them to summary metric table.
CCTV: figure out how to implement sweep
Write a dateflow diagram of what we gonna with DNV image. Get image , valid image, train, how we evaluate the result how I am going to validate images with access data. From video to train model
PDA vs DDA from system flow balance
ask deven if he has any question regarding time-series multiple variables
re-labeling image that does not have text as defect if the defect is invisible in the image. (do a spatial correlation with image with text)

Collimator: need to rename all the block. Send the list of block name to Pavan to review. Follow the nameing standard and python standard. See collimator.md for more information
CCTV: I submitted three jobs to compute canada after testing the code with sample dataset.
- run fastai for multi defect classification with image aug Resize((576, 720),method='squish', pad_mode='zeros'). I include method and pad_mod in the argument because if they are not included crop and reflection will be chosen as default method and pad_mode. The input dataset are 14 training 2 validation SewerML images folders.
- run fastai for multi defect classification with image aug Resize((576, 720),method='squish', pad_mode='zeros'). The job is run using only 14 training SewerML folders. The dataset is split into 80/20 for training and validation.
- run fastai for single water level classification using Danish 2015 standard with image aug Resize((576, 720),method='squish', pad_mode='zeros'). The input dataset are from 14 training and 2 validation SewerML images folders.
All the three jobs are still in the queue as of Sunday October 12.