EDIT
I was able to modify the pixel Points coordinates by using the rasterio.transform method, the SHP file works on QGIS, and the coordinates are almost aligned, but not exactly.
I tried many combination but never the shape exactly cover the TIFF extent...

import rasterio
from rasterio.plot import show
from rasterio.windows import Window
import shapefile


flat_list = [item for sublist in B for item in sublist]


w = shapefile.Writer('03-c')
w.field('name', 'C')

image = '/content/drive/My Drive/ML/03.tif'
with rasterio.open(image) as src:
    
    for i in range(len(flat_list)):
      ### X & Y is the center of the rectangle in flatlist[]  
      X = (int((flat_list[i][0])/2+flat_list[i][2]))
      Y = (src.shape[0] - int((flat_list[i][1]+flat_list[i][3])/2))
      Point = src.transform * (X, Y)
      w.point(Point[0], Point[1])
      w.record('linestring1')
    w.close()
  

I was able to modify the pixel Points coordinates by using the rasterio.transform method, the SHP file works on QGIS, and the coordinates are almost aligned, but not exactly.
I tried many combination but never the shape exactly cover the TIFF extent...

import rasterio
from rasterio.plot import show
from rasterio.windows import Window
import shapefile

flat_list = [item for sublist in B for item in sublist]

w = shapefile.Writer('03-c')
w.field('name', 'C')

image = '/content/drive/My Drive/ML/03.tif'
with rasterio.open(image) as src:
    
    for i in range(len(flat_list)):
      ### X & Y is the center of the rectangle in flatlist[]  
      X = (int((flat_list[i][0])/2+flat_list[i][2]))
      Y = (src.shape[0] - int((flat_list[i][1]+flat_list[i][3])/2))
      Point = src.transform * (X, Y)
      w.point(Point[0], Point[1])
      w.record('linestring1')
    w.close()

I made a point shapefile (using PySHP in Python) that represent objects detected through AI.
This shapefile has no CRS and each point is X,Y relative to the image it was processed from (actually a TIFF converted to JPG)
ex : point1 = (200,300) ; point2 = (600, 500)

I have the original TIFF and I can extract its information using some Python library (Rasterio):
tiff shape : (7579, 13887)
EPSG:4326
BoundingBox(left=-73.3363245134031, bottom=45.18278294324435, right=-73.33487807770662, top=45.18357239348942)

What is the correct process to build the point shapefile in order for its extent to be the same as the original TIFF?
-Should I modify the EPSG and boundaries of the shapefile so it recover the TIFF?
-Should I reproject each point while creating the SHP using pySHP? It seems this is the legitimate method, but how can I recalculate each point coordinates knowing only the EPSG and the boundaries of my TIFF?

I made a point shapefile (using PySHP in Python) that represent objects detected through AI.
This shapefile has no CRS and each point is X,Y relative to the image it was processed from (actually a TIFF converted to JPG)
ex : point1 = (200,300) ; point2 = (600, 500)

I have the original TIFF and I can extract its information using some Python library (Rasterio):
tiff shape : (7579, 13887)
EPSG:4326
BoundingBox(left=-73.3363245134031, bottom=45.18278294324435, right=-73.33487807770662, top=45.18357239348942)

What is the correct process to build the point shapefile in order for its extent to be the same as the original TIFF?
-Should I modify the EPSG and boundaries of the shapefile so it recover the TIFF?
-Should I reproject each point while creating the SHP using pySHP? It seems this is the legitimate method, but how can I recalculate each point coordinates knowing only the EPSG and the boundaries of my TIFF?

EDIT
I was able to modify the pixel Points coordinates by using the rasterio.transform method, the SHP file works on QGIS, and the coordinates are almost aligned, but not exactly.
I tried many combination but never the shape exactly cover the TIFF extent...

import rasterio
from rasterio.plot import show
from rasterio.windows import Window
import shapefile


flat_list = [item for sublist in B for item in sublist]


w = shapefile.Writer('03-c')
w.field('name', 'C')

image = '/content/drive/My Drive/ML/03.tif'
with rasterio.open(image) as src:
    
    for i in range(len(flat_list)):
      ### X & Y is the center of the rectangle in flatlist[]  
      X = (int((flat_list[i][0])/2+flat_list[i][2]))
      Y = (src.shape[0] - int((flat_list[i][1]+flat_list[i][3])/2))
      Point = src.transform * (X, Y)
      w.point(Point[0], Point[1])
      w.record('linestring1')
    w.close()
  

I made a point shapefile (using PySHP in Python) that represent objects detected through AI.
This shapefile has no CRS and each point is X,Y relative to the image it was processed from (actually a TIFF converted to JPG)
ex : point1 = (200,300) ; point2 = (600, 500)

I have the original TIFF and I can extract its information using some python librairy (Rasterio):
tiff shape : (7579, 13887)
EPSG:4326
BoundingBox(left=-73.3363245134031, bottom=45.18278294324435, right=-73.33487807770662, top=45.18357239348942)

What is the correct process to build the point shapefile in order for its extent to be the same as the original TIFF  ?
-Should I modify the EPSG and boundaries of the shapefile so it recover the TIFF  ?
-Should I reproject each point while creating the SHP using pySHP  ? It seems this is the legitimate method, but how can I recalculate each point coordinates knowing only the EPSG and the boundaries of my TIFF  ?

Thank you

I made a point shapefile (using PySHP in Python) that represent objects detected through AI.
This shapefile has no CRS and each point is X,Y relative to the image it was processed from (actually a TIFF converted to JPG)
ex : point1 = (200,300) ; point2 = (600, 500)

I have the original TIFF and I can extract its information using some Python library (Rasterio):
tiff shape : (7579, 13887)
EPSG:4326
BoundingBox(left=-73.3363245134031, bottom=45.18278294324435, right=-73.33487807770662, top=45.18357239348942)

What is the correct process to build the point shapefile in order for its extent to be the same as the original TIFF?
-Should I modify the EPSG and boundaries of the shapefile so it recover the TIFF?
-Should I reproject each point while creating the SHP using pySHP? It seems this is the legitimate method, but how can I recalculate each point coordinates knowing only the EPSG and the boundaries of my TIFF?