Commit 5ad70432 authored by Sabyasachi Mondal's avatar Sabyasachi Mondal
Browse files

commit: comparison label added , large size images tested , final readme updates

parent 28d018c1
This source diff could not be displayed because it is too large. You can view the blob instead.
This diff is collapsed.
File added
File added
File added
File added
File added
File added
......@@ -4,9 +4,10 @@
"cell_type": "markdown",
"metadata": {},
"source": [
"## Hyperspectral Vision Filter for satellite / birdseye view drone shots\n",
"Streaming in pixel data in 3xRow format to be fed to filter kernel and produce an edgemap the filter captures the Features in an Image\n",
"Application of Robert's convolution can be feature detection in areial mapping"
"## Image Filter for satellite images (single band IR)\n",
"Streaming in pixel data in 3xRow format to be fed to filter kernel and produce an edgemap the filter captures the Features in an Image.\n",
"Application of Robert's convolution can be feature detection in areial mapping and terrain feature detection.\n",
"Check the Notebook on comparison for CPU vs FPGA comparison https://mygit.th-deg.de/sm11312/fpga_final_project/-/blob/main/Notebook_Speed_Comparison.ipynb ."
]
},
{
......@@ -62,6 +63,13 @@
"import time"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Converting the whole image and checking the size of the stream"
]
},
{
"cell_type": "code",
"execution_count": 13,
......@@ -105,6 +113,13 @@
"plt.imshow(arr_2d)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Loading overlay"
]
},
{
"cell_type": "code",
"execution_count": 15,
......@@ -115,6 +130,13 @@
"oly.download()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Linking Python objects to DMAs"
]
},
{
"cell_type": "code",
"execution_count": 16,
......@@ -130,6 +152,13 @@
"dma4 = oly.OP_Channel1"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Looping to read the image row by row , 3 rows at a time , where each row is a DMA stream"
]
},
{
"cell_type": "code",
"execution_count": 17,
......@@ -179,6 +208,20 @@
" del op_ch1"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Output from 2 DMA channels (2 rows of the image) packed up as a numpy array along the height as the loop progress"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Checking output array width (2 less becuase convolution takes 2 streams and generate 1 so the ones at rightmost and leftmost are left out)"
]
},
{
"cell_type": "code",
"execution_count": 19,
......@@ -199,6 +242,13 @@
"len(out_arr[0])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Output array size (height of our image - 2) 2 streams again left out as explained above"
]
},
{
"cell_type": "code",
"execution_count": 20,
......@@ -257,6 +307,13 @@
"img_robert = (np.array(out_arr))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Converting the 16bit data to 8bit , the datastreams from DMA are 16 bit and little Endian but image pixels are 8 bit Little Endian"
]
},
{
"cell_type": "code",
"execution_count": 24,
......@@ -360,6 +417,13 @@
"ver = []"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#### Implementing same algorithm in CPU as in FPGA"
]
},
{
"cell_type": "code",
"execution_count": 29,
......
This diff is collapsed.
Output.jpg

46.8 KB | W: | H:

Output.jpg

240 KB | W: | H:

Output.jpg
Output.jpg
Output.jpg
Output.jpg
  • 2-up
  • Swipe
  • Onion skin
Output_cpu.jpg

39.8 KB | W: | H:

Output_cpu.jpg

130 KB | W: | H:

Output_cpu.jpg
Output_cpu.jpg
Output_cpu.jpg
Output_cpu.jpg
  • 2-up
  • Swipe
  • Onion skin
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment